[client] Move metrics saver into its own metricsstages package

Relocate MetricsStages (per-connection metric stage timestamps) out of
the peer package into client/internal/peer/metricsstages.

.coderabbit.yaml

@@ -0,0 +1,18 @@
+# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
+language: en-US
+reviews:
+  profile: chill
+  request_changes_workflow: false
+  high_level_summary: true
+  poem: false
+  review_status: true
+  auto_review:
+    enabled: true
+    drafts: false
+  path_filters:
+    - "!**/*.tsx"
+    - "!**/*.ts"
+    - "!**/*.js"
+    - "!**/*.svg"
+chat:
+  auto_reply: true

.devcontainer/Dockerfile

@@ -6,7 +6,6 @@ RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
         iptables=1.8.9-2 \
         libgl1-mesa-dev=22.3.6-1+deb12u1 \
         xorg-dev=1:7.7+23 \
-        libayatana-appindicator3-dev=0.5.92-1 \
     && apt-get clean \
     && rm -rf /var/lib/apt/lists/* \
     && go install -v golang.org/x/tools/gopls@latest

.github/workflows/frontend-ui.yml

@@ -0,0 +1,98 @@
+name: UI Frontend
+
+on:
+  pull_request:
+    paths:
+      - "client/ui/frontend/**"
+      - "client/ui/i18n/**"
+      - "client/ui/**/*.go"
+      - ".github/workflows/frontend-ui.yml"
+  push:
+    branches:
+      - main
+    paths:
+      - "client/ui/frontend/**"
+      - "client/ui/i18n/**"
+      - "client/ui/**/*.go"
+
+permissions:
+  contents: read
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref || github.actor_id }}
+  cancel-in-progress: true
+
+jobs:
+  lint-and-build:
+    name: Lint & Build
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+    defaults:
+      run:
+        working-directory: client/ui/frontend
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+
+      - name: Set up pnpm
+        uses: pnpm/action-setup@a3252b78c470c02df07e9d59298aecedc3ccdd6d # v3.0.0
+        with:
+          version: 11
+
+      # Bindings are generated by wails3 from the Go service definitions and
+      # are not checked in (see client/ui/frontend/bindings/). Without them,
+      # typecheck/build fail on missing module imports.
+      - name: Set up Go
+        uses: actions/setup-go@4b73464bb391d4059bd26b0524d20df3927bd417 # v6.3.0
+        with:
+          go-version-file: "go.mod"
+          cache: false
+
+      # wails3 CLI links against GTK4 / WebKitGTK 6.0 via its internal/operatingsystem
+      # package, so the dev libraries must be present before `go install`.
+      - name: Install Wails Linux system dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends \
+            pkg-config \
+            libgtk-4-dev \
+            libwebkitgtk-6.0-dev
+
+      - name: Install wails3 CLI
+        # Version derived from go.mod so the binding generator always matches
+        # the wails runtime the daemon links against.
+        working-directory: ${{ github.workspace }}
+        run: |
+          WAILS_VERSION=$(go list -m -f '{{.Version}}' github.com/wailsapp/wails/v3)
+          go install github.com/wailsapp/wails/v3/cmd/wails3@$WAILS_VERSION
+
+      - name: Get pnpm store directory
+        id: pnpm-store
+        run: echo "path=$(pnpm store path --silent)" >> "$GITHUB_OUTPUT"
+
+      - name: Cache pnpm store
+        uses: actions/cache@v4
+        with:
+          path: ${{ steps.pnpm-store.outputs.path }}
+          key: ${{ runner.os }}-pnpm-${{ hashFiles('client/ui/frontend/pnpm-lock.yaml') }}
+          restore-keys: |
+            ${{ runner.os }}-pnpm-
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Generate Wails bindings
+        run: pnpm run bindings
+
+      - name: Lint, typecheck, format
+        run: pnpm check
+
+      - name: Build
+        run: pnpm build

.github/workflows/golang-test-darwin.yml

@@ -45,7 +45,15 @@ jobs:
         run: git --no-pager diff --exit-code
 
       - name: Test
-        run: NETBIRD_STORE_ENGINE=${{ matrix.store }} CI=true go test -coverprofile=coverage.txt -tags 'devcert privileged' -exec 'sudo --preserve-env=CI,NETBIRD_STORE_ENGINE' -timeout 5m -p 1 $(go list ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/testutil/privileged)
+        # Exclude client/ui: its main.go uses //go:embed all:frontend/dist,
+        # which fails to compile until the frontend has been built. The Wails UI
+        # has no Go-side unit tests, and its release pipeline runs `pnpm build`
+        # before goreleaser.
+        # `go list -e` lets the listing succeed even though the embed fails to
+        # resolve; the grep then drops the broken package by path. Without -e,
+        # go list aborts with empty stdout and `go test` falls back to the repo
+        # root, which has no Go files.
+        run: NETBIRD_STORE_ENGINE=${{ matrix.store }} CI=true go test -coverprofile=coverage.txt -tags 'devcert privileged' -exec 'sudo --preserve-env=CI,NETBIRD_STORE_ENGINE' -timeout 5m -p 1 $(go list -e ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/ui -e /client/testutil/privileged)
 
       - name: Upload coverage reports to Codecov
         uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f #v7.0.0

.github/workflows/golang-test-linux.yml

@@ -53,7 +53,7 @@ jobs:
 
       - name: Install dependencies
         if: steps.cache.outputs.cache-hit != 'true'
-        run: sudo apt update && sudo apt install -y -q libgtk-3-dev libayatana-appindicator3-dev libgl1-mesa-dev xorg-dev gcc-multilib libpcap-dev
+        run: sudo apt update && sudo apt install -y -q libgtk-4-dev libwebkitgtk-6.0-dev libsoup-3.0-dev libgl1-mesa-dev xorg-dev gcc-multilib libpcap-dev
 
       - name: Install 32-bit libpcap
         if: steps.cache.outputs.cache-hit != 'true'
@@ -145,7 +145,7 @@ jobs:
             ${{ runner.os }}-gotest-cache-
 
       - name: Install dependencies
-        run: sudo apt update && sudo apt install -y -q libgtk-3-dev libayatana-appindicator3-dev libgl1-mesa-dev xorg-dev gcc-multilib libpcap-dev
+        run: sudo apt update && sudo apt install -y -q libgtk-4-dev libwebkitgtk-6.0-dev libsoup-3.0-dev libgl1-mesa-dev xorg-dev gcc-multilib libpcap-dev
 
       - name: Install 32-bit libpcap
         if: matrix.arch == '386'
@@ -158,7 +158,15 @@ jobs:
         run: git --no-pager diff --exit-code
 
       - name: Test
-        run: CGO_ENABLED=1 GOARCH=${{ matrix.arch }} CI=true go test -coverprofile=coverage.txt -tags 'devcert privileged' -exec 'sudo --preserve-env=CI,CGO_ENABLED' -timeout 10m -p 1 $(go list ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/testutil/privileged)
+        # Exclude client/ui: its main.go uses //go:embed all:frontend/dist,
+        # which fails to compile until the frontend has been built. The Wails UI
+        # has no Go-side unit tests, and its release pipeline runs `pnpm build`
+        # before goreleaser.
+        # `go list -e` lets the listing succeed even though the embed fails to
+        # resolve; the grep then drops the broken package by path. Without -e,
+        # go list aborts with empty stdout and `go test` falls back to the repo
+        # root, which has no Go files.
+        run: CGO_ENABLED=1 GOARCH=${{ matrix.arch }} CI=true go test -coverprofile=coverage.txt -tags 'devcert privileged' -exec 'sudo --preserve-env=CI,CGO_ENABLED' -timeout 10m -p 1 $(go list -e ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/ui -e /client/testutil/privileged)
 
       - name: Upload coverage reports to Codecov
         if: matrix.arch == 'amd64'
@@ -168,7 +176,6 @@ jobs:
           slug: netbirdio/netbird
           flags: unit,client
 
-
   test_client_on_docker:
     name: "Client (Docker) / Unit"
     needs: [build-cache]
@@ -229,7 +236,7 @@ jobs:
             sh -c ' \
               apk update; apk add --no-cache \
                 ca-certificates iptables ip6tables dbus dbus-dev libpcap-dev build-base; \
-              go test -buildvcs=false -tags "devcert privileged" -v -timeout 10m -p 1 $(go list -buildvcs=false ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/ui -e /upload-server -e /client/testutil/privileged)
+              go test -buildvcs=false -tags "devcert privileged" -v -timeout 10m -p 1 $(go list -e -buildvcs=false ./... | grep -v -e /management -e /signal -e /relay -e /proxy -e /combined -e /client/ui -e /upload-server -e /client/testutil/privileged)
             '
 
   test_relay:

.github/workflows/golang-test-windows.yml

@@ -65,8 +65,15 @@ jobs:
       - run: PsExec64 -s -w ${{ github.workspace }} C:\hostedtoolcache\windows\go\${{ steps.go.outputs.go-version }}\x64\bin\go.exe env -w GOCACHE=${{ env.modcache }}
       - run: PsExec64 -s -w ${{ github.workspace }} C:\hostedtoolcache\windows\go\${{ steps.go.outputs.go-version }}\x64\bin\go.exe mod tidy
       - name: Generate test script
+        # Exclude client/ui: its main.go uses //go:embed all:frontend/dist,
+        # which fails to compile until the frontend has been built. The Wails UI
+        # has no Go-side unit tests, and its release pipeline runs `pnpm build`
+        # before goreleaser.
+        # `go list -e` lets the listing succeed even though the embed fails to
+        # resolve; the Where-Object pipeline then drops the broken package by
+        # path. Without -e, go list aborts with empty stdout.
         run: |
-          $packages = go list ./... | Where-Object { $_ -notmatch '/management' } | Where-Object { $_ -notmatch '/relay' } | Where-Object { $_ -notmatch '/signal' } | Where-Object { $_ -notmatch '/proxy' } | Where-Object { $_ -notmatch '/combined' }
+          $packages = go list -e ./... | Where-Object { $_ -notmatch '/management' } | Where-Object { $_ -notmatch '/relay' } | Where-Object { $_ -notmatch '/signal' } | Where-Object { $_ -notmatch '/proxy' } | Where-Object { $_ -notmatch '/combined' } | Where-Object { $_ -notmatch '/client/ui' }
           $goExe = "C:\hostedtoolcache\windows\go\${{ steps.go.outputs.go-version }}\x64\bin\go.exe"
           $cmd = "$goExe test -tags `"devcert privileged`" -timeout 10m -p 1 $($packages -join ' ') > test-out.txt 2>&1"
           Set-Content -Path "${{ github.workspace }}\run-tests.cmd" -Value $cmd

.github/workflows/golangci-lint.yml

@@ -22,7 +22,15 @@ jobs:
         uses: codespell-project/actions-codespell@8f01853be192eb0f849a5c7d721450e7a467c579 # v2.2
         with:
           ignore_words_list: erro,clienta,hastable,iif,groupd,testin,groupe,cros,ans,deriver,te,userA,ede,additionals,flate,recordin,unparseable
-          skip: go.mod,go.sum,**/proxy/web/**
+          # Non-English UI translations trip codespell on real foreign words
+          # (de: "Sie", "oder", "ist"). Only en/common.json is the source of
+          # truth that should be spell-checked. List each translated locale
+          # dir below and add new ones as languages are added under
+          # client/ui/i18n/locales/. Single-star globs are matched per path
+          # segment by codespell and behave the same across versions; the
+          # recursive "**" form did not take effect with the codespell shipped
+          # by this action.
+          skip: go.mod,go.sum,*/proxy/web/*,*pnpm-lock.yaml,*package-lock.json,*/locales/de/*,*/locales/es/*,*/locales/fr/*,*/locales/hu/*,*/locales/it/*,*/locales/pt/*,*/locales/ru/*,*/locales/zh-CN/*,*/i18n/TRANSLATING.md
   golangci:
     strategy:
       fail-fast: false
@@ -54,7 +62,16 @@ jobs:
           cache: false
       - name: Install dependencies
         if: matrix.os == 'ubuntu-latest'
-        run: sudo apt update && sudo apt install -y -q libgtk-3-dev libayatana-appindicator3-dev libgl1-mesa-dev xorg-dev libpcap-dev
+        run: sudo apt update && sudo apt install -y -q libgtk-4-dev libwebkitgtk-6.0-dev libsoup-3.0-dev libgl1-mesa-dev xorg-dev libpcap-dev
+      - name: Stub Wails frontend bundle
+        # client/ui/main.go has //go:embed all:frontend/dist. The
+        # directory is produced by `pnpm run build` and is gitignored, so
+        # lint-only runs (no frontend toolchain) need a placeholder file
+        # for the embed pattern to match.
+        shell: bash
+        run: |
+          mkdir -p client/ui/frontend/dist
+          touch client/ui/frontend/dist/.embed-placeholder
       - name: golangci-lint
         uses: golangci/golangci-lint-action@82606bf257cbaff209d206a39f5134f0cfbfd2ee #v9.2.1
         with:

.github/workflows/release.yml

@@ -9,7 +9,7 @@ on:
   pull_request:
 
 env:
-  SIGN_PIPE_VER: "v0.1.6"
+  SIGN_PIPE_VER: "v0.1.8"
   GORELEASER_VER: "v2.16.0"
   PRODUCT_NAME: "NetBird"
   COPYRIGHT: "NetBird GmbH"
@@ -216,9 +216,9 @@ jobs:
       - name: Install goversioninfo
         run: go install github.com/josephspurrier/goversioninfo/cmd/goversioninfo@233067e
       - name: Generate windows syso amd64
-        run: goversioninfo  -icon client/ui/assets/netbird.ico -manifest client/manifest.xml -product-name ${{ env.PRODUCT_NAME }} -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/resources_windows_amd64.syso
+        run: goversioninfo  -icon client/ui/build/windows/icon.ico -manifest client/manifest.xml -product-name ${{ env.PRODUCT_NAME }} -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/resources_windows_amd64.syso
       - name: Generate windows syso arm64
-        run: goversioninfo -arm -64 -icon client/ui/assets/netbird.ico -manifest client/manifest.xml -product-name ${{ env.PRODUCT_NAME }} -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/resources_windows_arm64.syso
+        run: goversioninfo -arm -64 -icon client/ui/build/windows/icon.ico -manifest client/manifest.xml -product-name ${{ env.PRODUCT_NAME }} -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/resources_windows_arm64.syso
       - name: Run GoReleaser
         id: goreleaser
         uses: goreleaser/goreleaser-action@5daf1e915a5f0af01ddbcd89a43b8061ff4f1a89 # v7.2.2
@@ -397,8 +397,18 @@ jobs:
       - name: check git status
         run: git --no-pager diff --exit-code
 
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: '22'
+
+      - name: Set up pnpm
+        uses: pnpm/action-setup@a3252b78c470c02df07e9d59298aecedc3ccdd6d # v3.0.0
+        with:
+          version: 11
+
       - name: Install dependencies
-        run: sudo apt update && sudo apt install -y -q libappindicator3-dev gir1.2-appindicator3-0.1 libxxf86vm-dev gcc-mingw-w64-x86-64
+        run: sudo apt update && sudo apt install -y -q libgtk-4-dev libwebkitgtk-6.0-dev libsoup-3.0-dev gcc-mingw-w64-x86-64
 
       - name: Decode GPG signing key
         if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
@@ -417,10 +427,16 @@ jobs:
           echo "/tmp/llvm-mingw-20250709-ucrt-ubuntu-22.04-x86_64/bin" >> $GITHUB_PATH
       - name: Install goversioninfo
         run: go install github.com/josephspurrier/goversioninfo/cmd/goversioninfo@233067e
+      - name: Install wails3 CLI
+        # Version derived from go.mod so the binding generator always matches
+        # the wails runtime the binary links against.
+        run: |
+          WAILS_VERSION=$(go list -m -f '{{.Version}}' github.com/wailsapp/wails/v3)
+          go install github.com/wailsapp/wails/v3/cmd/wails3@$WAILS_VERSION
       - name: Generate windows syso amd64
-        run: goversioninfo -64 -icon client/ui/assets/netbird.ico -manifest client/ui/manifest.xml -product-name ${{ env.PRODUCT_NAME }}-"UI" -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/ui/resources_windows_amd64.syso
+        run: goversioninfo -64 -icon client/ui/build/windows/icon.ico -manifest client/ui/build/windows/wails.exe.manifest -product-name ${{ env.PRODUCT_NAME }}-"UI" -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/ui/resources_windows_amd64.syso
       - name: Generate windows syso arm64
-        run: goversioninfo -arm -64 -icon client/ui/assets/netbird.ico -manifest client/ui/manifest.xml -product-name ${{ env.PRODUCT_NAME }}-"UI" -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/ui/resources_windows_arm64.syso
+        run: goversioninfo -arm -64 -icon client/ui/build/windows/icon.ico -manifest client/ui/build/windows/wails.exe.manifest -product-name ${{ env.PRODUCT_NAME }}-"UI" -copyright "${{ env.COPYRIGHT }}" -ver-major ${{ steps.semver_parser.outputs.major }} -ver-minor ${{ steps.semver_parser.outputs.minor }} -ver-patch ${{ steps.semver_parser.outputs.patch }} -ver-build 0 -file-version ${{ steps.semver_parser.outputs.fullversion }}.0 -product-version ${{ steps.semver_parser.outputs.fullversion }}.0 -o client/ui/resources_windows_arm64.syso
 
       - name: Run GoReleaser
         uses: goreleaser/goreleaser-action@5daf1e915a5f0af01ddbcd89a43b8061ff4f1a89 # v7.2.2
@@ -489,6 +505,20 @@ jobs:
         run: go mod tidy
       - name: check git status
         run: git --no-pager diff --exit-code
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: '22'
+      - name: Set up pnpm
+        uses: pnpm/action-setup@a3252b78c470c02df07e9d59298aecedc3ccdd6d # v3.0.0
+        with:
+          version: 11
+      - name: Install wails3 CLI
+        # Version derived from go.mod so the binding generator always matches
+        # the wails runtime the binary links against.
+        run: |
+          WAILS_VERSION=$(go list -m -f '{{.Version}}' github.com/wailsapp/wails/v3)
+          go install github.com/wailsapp/wails/v3/cmd/wails3@$WAILS_VERSION
       - name: Run GoReleaser
         id: goreleaser
         uses: goreleaser/goreleaser-action@5daf1e915a5f0af01ddbcd89a43b8061ff4f1a89 # v7.2.2
@@ -576,23 +606,6 @@ jobs:
       - name: Move wintun.dll into dist
         run: mv ${{ env.downloadPath }}\wintun\bin\${{ matrix.wintun_arch }}\wintun.dll ${{ github.workspace }}\dist\${{ env.PackageWorkdir }}\
 
-      - name: Download Mesa3D (amd64 only)
-        id: download-mesa3d
-        if: matrix.arch == 'amd64'
-        uses: netbirdio/shared-actions/actions/win-download-and-verify@be5df6047383da2236e02243cceb857d8567c27e # v0.0.2
-        with:
-          url: https://pkgs.netbird.io/mesa3d/MesaForWindows-x64-20.1.8.7z
-          destination: ${{ env.downloadPath }}\mesa3d.7z
-          sha256: 71c7cb64ec229a1d6b8d62fa08e1889ed2bd17c0eeede8689daf0f25cb31d6b9
-
-      - name: Extract Mesa3D driver (amd64 only)
-        if: matrix.arch == 'amd64'
-        run: 7z x -o"${{ env.downloadPath }}" "${{ env.downloadPath }}/mesa3d.7z"
-
-      - name: Move opengl32.dll into dist (amd64 only)
-        if: matrix.arch == 'amd64'
-        run: mv ${{ env.downloadPath }}\opengl32.dll ${{ github.workspace }}\dist\${{ env.PackageWorkdir }}\
-
       - name: Download EnVar plugin for NSIS
         uses: netbirdio/shared-actions/actions/win-download-and-verify@be5df6047383da2236e02243cceb857d8567c27e # v0.0.2
         with:
@@ -615,6 +628,28 @@ jobs:
         if: matrix.arch == 'amd64'
         run: 7z x -o"${{ github.workspace }}/NSIS_Plugins" "${{ github.workspace }}/ShellExecAsUser_amd64-Unicode.7z"
 
+      - name: Set up Go for wails3 CLI
+        uses: actions/setup-go@v5
+        with:
+          go-version-file: "go.mod"
+          cache: false
+
+      - name: Install wails3 CLI
+        # Version derived from go.mod so the bootstrapper payload always
+        # matches the wails runtime the binary links against.
+        shell: bash
+        run: |
+          WAILS_VERSION=$(go list -m -f '{{.Version}}' github.com/wailsapp/wails/v3)
+          go install github.com/wailsapp/wails/v3/cmd/wails3@$WAILS_VERSION
+
+      - name: Stage WebView2 bootstrapper for installers
+        # Both client/installer.nsis and client/netbird.wxs reference
+        # client/MicrosoftEdgeWebview2Setup.exe. wails3 writes it there.
+        # The signing pipeline (netbirdio/sign-pipelines) does the same
+        # step for release builds; this mirrors it for PR sanity testing.
+        shell: bash
+        run: wails3 generate webview2bootstrapper -dir client
+
       - name: Build NSIS installer
         shell: pwsh
         env:

.github/workflows/wasm-build-validation.yml

@@ -27,7 +27,7 @@ jobs:
         with:
           go-version-file: "go.mod"
       - name: Install dependencies
-        run: sudo apt update && sudo apt install -y -q libgtk-3-dev libayatana-appindicator3-dev libgl1-mesa-dev xorg-dev libpcap-dev
+        run: sudo apt update && sudo apt install -y -q libgtk-4-dev libwebkitgtk-6.0-dev libsoup-3.0-dev libgl1-mesa-dev xorg-dev libpcap-dev
       - name: Install golangci-lint
         uses: golangci/golangci-lint-action@82606bf257cbaff209d206a39f5134f0cfbfd2ee #v9.2.1
         with:

.golangci.yaml

@@ -114,6 +114,16 @@ linters:
       - linters:
           - staticcheck
         text: "QF1012"
+      # client/ui/main.go uses //go:embed all:frontend/dist; the
+      # directory is populated by `pnpm build` in the release pipeline
+      # and missing at lint time, so the embed parses to "no matching
+      # files found" — surfaced by golangci-lint's typecheck pre-pass.
+      # Suppress just that one diagnostic; the rest of the package
+      # (services/, tray.go, grpc.go, ...) still gets linted normally.
+      - linters:
+          - typecheck
+        path: client/ui/main\.go
+        text: "pattern all:frontend/dist"
     paths:
       - third_party$
       - builtin$

.goreleaser.yaml

@@ -212,6 +212,7 @@ nfpms:
     description: Netbird client.
     homepage: https://netbird.io/
     license: BSD-3-Clause
+    vendor: NetBird
     id: netbird_deb
     bindir: /usr/bin
     builds:
@@ -226,6 +227,7 @@ nfpms:
     description: Netbird client.
     homepage: https://netbird.io/
     license: BSD-3-Clause
+    vendor: NetBird
     id: netbird_rpm
     bindir: /usr/bin
     builds:

.goreleaser_ui.yaml

@@ -2,6 +2,15 @@ version: 2
 env:
   - SKIP_PUBLISH={{ if index .Env "SKIP_PUBLISH" }}{{ .Env.SKIP_PUBLISH }}{{ else }}true{{ end }}
 project_name: netbird-ui
+
+before:
+  hooks:
+    # Bindings are gitignored; regenerate before the frontend build so
+    # the @wailsio/runtime Vite plugin can resolve them (vite refuses to
+    # build without them).
+    - sh -c 'cd client/ui && wails3 generate bindings -clean=true -ts'
+    - sh -c 'cd client/ui/frontend && pnpm install --frozen-lockfile && pnpm build'
+
 builds:
   - id: netbird-ui
     dir: client/ui
@@ -62,6 +71,8 @@ nfpms:
   - maintainer: Netbird <dev@netbird.io>
     description: Netbird client UI.
     homepage: https://netbird.io/
+    license: BSD-3-Clause
+    vendor: NetBird
     id: netbird_ui_deb
     package_name: netbird-ui
     builds:
@@ -71,9 +82,9 @@ nfpms:
     scripts:
       postinstall: "release_files/ui-post-install.sh"
     contents:
-      - src: client/ui/build/netbird.desktop
-        dst: /usr/share/applications/netbird.desktop
-      - src: client/ui/assets/netbird.png
+      - src: client/ui/build/linux/netbird.desktop
+        dst: /usr/share/applications/org.wails.netbird.desktop
+      - src: client/ui/build/appicon.png
         dst: /usr/share/pixmaps/netbird.png
     dependencies:
       - netbird
@@ -81,6 +92,8 @@ nfpms:
   - maintainer: Netbird <dev@netbird.io>
     description: Netbird client UI.
     homepage: https://netbird.io/
+    license: BSD-3-Clause
+    vendor: NetBird
     id: netbird_ui_rpm
     package_name: netbird-ui
     builds:
@@ -90,12 +103,13 @@ nfpms:
     scripts:
       postinstall: "release_files/ui-post-install.sh"
     contents:
-      - src: client/ui/build/netbird.desktop
-        dst: /usr/share/applications/netbird.desktop
-      - src: client/ui/assets/netbird.png
+      - src: client/ui/build/linux/netbird.desktop
+        dst: /usr/share/applications/org.wails.netbird.desktop
+      - src: client/ui/build/appicon.png
         dst: /usr/share/pixmaps/netbird.png
     dependencies:
       - netbird
+
     rpm:
       signature:
         key_file: '{{ if index .Env "GPG_RPM_KEY_FILE" }}{{ .Env.GPG_RPM_KEY_FILE }}{{ end }}'

.goreleaser_ui_darwin.yaml

@@ -1,6 +1,15 @@
 version: 2
 
 project_name: netbird-ui
+
+before:
+  hooks:
+    # Bindings are gitignored; regenerate before the frontend build so
+    # the @wailsio/runtime Vite plugin can resolve them (vite refuses to
+    # build without them).
+    - sh -c 'cd client/ui && wails3 generate bindings -clean=true -ts'
+    - sh -c 'cd client/ui/frontend && pnpm install --frozen-lockfile && pnpm build'
+
 builds:
   - id: netbird-ui-darwin
     dir: client/ui
@@ -20,8 +29,6 @@ builds:
     ldflags:
       - -s -w -X github.com/netbirdio/netbird/version.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}} -X main.builtBy=goreleaser
     mod_timestamp: "{{ .CommitTimestamp }}"
-    tags:
-      - load_wgnt_from_rsrc
 
 universal_binaries:
   - id: netbird-ui-darwin

CONTRIBUTING.md

@@ -79,13 +79,21 @@ dependencies are installed. Here is a short guide on how that can be done.
 
 ### Requirements
 
-#### Go 1.21
+#### Go 1.25
 
 Follow the installation guide from https://go.dev/
 
-#### UI client - Fyne toolkit 
+#### UI client - Wails v3 + React
 
-We use the fyne toolkit in our UI client. You can follow its requirement guide to have all its dependencies installed: https://developer.fyne.io/started/#prerequisites
+The desktop UI client (`client/ui`) is built with [Wails v3](https://v3.wails.io/) and a React frontend rendered in a WebView. To build it you need:
+
+- Go ≥ 1.25
+- Node ≥ 20 and **pnpm** (`corepack enable && corepack prepare pnpm@latest --activate`)
+- The `wails3` CLI: `go install github.com/wailsapp/wails/v3/cmd/wails3@latest`
+- The `task` runner: `go install github.com/go-task/task/v3/cmd/task@latest`
+- Linux only: `libwebkitgtk-6.0-dev`, `libgtk-4-dev`, `libsoup-3.0-dev`
+
+All UI build, dev-loop, and cross-compile commands are described in the [UI client](#ui-client) section below.
 
 #### gRPC
 You can follow the instructions from the quickstarter guide https://grpc.io/docs/languages/go/quickstart/#prerequisites and then run the `generate.sh` files located in each `proto` directory to generate changes.
@@ -214,6 +222,39 @@ To start NetBird the client in the foreground:
 sudo ./client up --log-level debug --log-file console
 ```
 > On Windows use a powershell with administrator privileges
+
+#### UI client
+
+The desktop UI lives in `client/ui` and is built with Wails v3 (see [Requirements](#ui-client---wails-v3--react)). All commands run from `client/ui`.
+
+Live-reload development (Vite + Go binary + `*.go` watcher):
+
+```
+cd client/ui
+task dev
+```
+
+Pass daemon flags after `--`:
+
+```
+task dev -- --daemon-addr=tcp://127.0.0.1:41731
+```
+
+Production build (frontend assets embedded into the binary, output in `client/ui/bin/`):
+
+```
+cd client/ui
+task build
+```
+
+Cross-compile the Windows binary from Linux (requires the mingw-w64 toolchain, e.g. `sudo apt install gcc-mingw-w64-x86-64`):
+
+```
+CGO_ENABLED=1 task windows:build
+```
+
+> macOS cross-compile from Linux is not supported (signing and notarization need a real Mac).
+
 #### Signal service
 
 To start NetBird's signal, execute:
@@ -251,10 +292,10 @@ Create dist directory
 mkdir -p dist/netbird_windows_amd64
 ```
 
-UI client
+UI client (built with Wails v3 — see the [UI client](#ui-client) section above)
 ```shell
-CC=x86_64-w64-mingw32-gcc CGO_ENABLED=1 GOOS=windows GOARCH=amd64 go build -o netbird-ui.exe -ldflags "-s -w -H windowsgui" ./client/ui
-mv netbird-ui.exe ./dist/netbird_windows_amd64/
+(cd client/ui && CGO_ENABLED=1 task windows:build)
+mv client/ui/bin/netbird-ui.exe ./dist/netbird_windows_amd64/
 ```
 
 Client
@@ -291,8 +332,6 @@ go test -exec sudo ./...
 ```
 > On Windows use a powershell with administrator privileges
 
-> Non-GTK environments will need the `libayatana-appindicator3-dev` (debian/ubuntu) package installed
-
 ## Checklist before submitting a PR
 As a critical network service and open-source project, we must enforce a few things before submitting the pull-requests:
 - Keep functions as simple as possible, with a single purpose

agent-network/README.md

@@ -1,16 +1,47 @@
 # NetBird Agent Network
 
-Agent Network is NetBird's access control layer for AI agents and the people who run
-them. It gives every agent a real identity, tied to your identity provider (IdP), and
-governs what it can reach — the LLM APIs and AI gateways it can call, and the internal
-resources it can access. Traffic flows only over the encrypted NetBird tunnel, scoped by
-policy, with no API keys to leak.
+Agent Network is NetBird's access control layer for AI agents and the people who run them. 
+It gives every agent a real identity, tied to an identity provider (IdP), and governs what it can reach: LLM APIs and 
+AI gateways it can call, and the internal resources it can access. Traffic flows only over the encrypted NetBird tunnel, 
+scoped by policy, with no API keys or other credentials to leak. It also gives you control over cost and token usage.
 
-> **Beta.** Agent Network is open source and can be self-hosted on your own
-> infrastructure.
+Because every LLM request passes through an
+identity-aware proxy, you can:
+
+- **Set spending and rate limits** per agent, per user, or per team — with hard caps
+  that stop requests once a budget is reached.
+- **Restrict models and providers** so agents can only call approved (and cost-appropriate)
+  endpoints, keeping expensive models off-limits unless explicitly allowed.
+- **Attribute usage** by tracking token consumption and cost per identity, group, or cost center so every
+  request is tied back to the agent and person responsible.
+- **Reuse your existing AI gateway** — point the proxy at a gateway you already run,
+  keeping its routing and config in place while it adds identity on top, so you skip
+  API key distribution.
+
+https://github.com/user-attachments/assets/44d18286-d8ab-49f8-a457-98ccd66f3268
+
+> **Beta.** Agent Network is in beta, but it's stable and already running in
+> production environments. It's fully open source and can be self-hosted on your own
+> infrastructure, with no vendor lock-in and no data leaving your environment.
 
 ## How it works
 
+Say you have a simple use case: your Engineering or IT team needs access to Claude Code or Codex, and you want visibility into usage plus the ability to enforce budgets.
+How can you do that without creating a dedicated API key for every team?
+
+With Agent Network you get a private endpoint inside your network, for example: https://mirror.netbird.ai
+Teams configure their agents to point to that endpoint instead of using individual API keys directly. 
+
+This endpoint is only reachable when users are connected to your NetBird network and authenticated through your IdP. Otherwise, it is not accessible from the public internet.
+You can then use this private endpoint to configure your AI agents, whether that is Claude Code, Codex, or another tool.
+
+## Quickstart
+
+Full step-by-step setup:
+**https://docs.netbird.io/agent-network/quickstart**
+
+## Architecture 
+
 Agent Network is built on two existing NetBird capabilities:
 
 - **Overlay network** — the encrypted WireGuard mesh between peers.
@@ -22,6 +53,9 @@ LLM traffic is routed through the proxy's identity-aware pipeline, while interna
 resources (databases, internal APIs, self-hosted models) are reached directly over
 peer-to-peer WireGuard tunnels, governed by the same identities and access policies.
 
+<img width="4720" height="2218" alt="image" src="https://github.com/user-attachments/assets/1afa5da1-4b82-4f8a-a7a8-f417efadf1eb" />
+
+
 ## Where the code lives
 
 There is no separate "agent-network" service — it reuses the reverse-proxy and management

client/cmd/login.go

@@ -22,11 +22,19 @@ import (
 	"github.com/netbirdio/netbird/util"
 )
 
+// extendSessionFlag drives the `netbird login --extend` flow: refresh the
+// SSO session expiry on the management server without tearing down the
+// tunnel. Mutually exclusive with setup-key login (a setup-key cannot
+// refresh an SSO-tracked peer — see auth.errSetupKeyOnSSOExpiredPeer).
+var extendSessionFlag bool
+
 func init() {
 	loginCmd.PersistentFlags().BoolVar(&noBrowser, noBrowserFlag, false, noBrowserDesc)
 	loginCmd.PersistentFlags().BoolVar(&showQR, showQRFlag, false, showQRDesc)
 	loginCmd.PersistentFlags().StringVar(&profileName, profileNameFlag, "", profileNameDesc)
 	loginCmd.PersistentFlags().StringVarP(&configPath, "config", "c", "", "(DEPRECATED) Netbird config file location")
+	loginCmd.PersistentFlags().BoolVar(&extendSessionFlag, "extend", false,
+		"refresh the SSO session expiry without tearing down the tunnel (requires an active connection)")
 }
 
 var loginCmd = &cobra.Command{
@@ -61,6 +69,16 @@ var loginCmd = &cobra.Command{
 			return err
 		}
 
+		if extendSessionFlag {
+			if providedSetupKey != "" {
+				return fmt.Errorf("--extend cannot be combined with a setup key; setup keys can only enrol new peers")
+			}
+			if err := doExtendSession(ctx, cmd); err != nil {
+				return fmt.Errorf("extend session failed: %v", err)
+			}
+			return nil
+		}
+
 		// workaround to run without service
 		if util.FindFirstLogPath(logFiles) == "" {
 			if err := doForegroundLogin(ctx, cmd, providedSetupKey, activeProf); err != nil {
@@ -152,6 +170,65 @@ func doDaemonLogin(ctx context.Context, cmd *cobra.Command, providedSetupKey str
 	return nil
 }
 
+// doExtendSession drives the daemon's RequestExtendAuthSession /
+// WaitExtendAuthSession pair. The user is sent through a regular SSO flow
+// (browser + verification URL) and the resulting JWT is forwarded to the
+// management server's ExtendAuthSession RPC. The tunnel stays up
+// throughout — no Down/Up, no network-map resync.
+func doExtendSession(ctx context.Context, cmd *cobra.Command) error {
+	conn, err := DialClientGRPCServer(ctx, daemonAddr)
+	if err != nil {
+		//nolint
+		return fmt.Errorf("failed to connect to daemon error: %v\n"+
+			"If the daemon is not running please run: "+
+			"\nnetbird service install \nnetbird service start\n", err)
+	}
+	defer conn.Close()
+
+	client := proto.NewDaemonServiceClient(conn)
+
+	req := &proto.RequestExtendAuthSessionRequest{}
+	// Pre-fill the IdP login hint from the active profile so the user
+	// doesn't have to retype their email. Best-effort: we still proceed
+	// without a hint if the lookup fails.
+	pm := profilemanager.NewProfileManager()
+	if active, perr := pm.GetActiveProfile(); perr == nil {
+		if profState, sperr := pm.GetProfileState(active.ID); sperr == nil && profState.Email != "" {
+			req.Hint = &profState.Email
+		}
+	}
+
+	startResp, err := client.RequestExtendAuthSession(ctx, req)
+	if err != nil {
+		return fmt.Errorf("start extend session: %v", err)
+	}
+
+	uri := startResp.GetVerificationURIComplete()
+	if uri == "" {
+		uri = startResp.GetVerificationURI()
+	}
+	openURL(cmd, uri, startResp.GetUserCode(), noBrowser, showQR)
+
+	waitResp, err := client.WaitExtendAuthSession(ctx, &proto.WaitExtendAuthSessionRequest{
+		DeviceCode: startResp.GetDeviceCode(),
+		UserCode:   startResp.GetUserCode(),
+	})
+	if err != nil {
+		return fmt.Errorf("wait for extend session: %v", err)
+	}
+
+	if ts := waitResp.GetSessionExpiresAt(); ts.IsValid() && !ts.AsTime().IsZero() {
+		deadline := ts.AsTime().Local()
+		cmd.Printf("Session extended. New expiry: %s\n", deadline.Format("2006-01-02 15:04:05 MST"))
+	} else {
+		// Management reported the peer is not eligible (e.g. login
+		// expiration disabled on the account). Surface that fact
+		// instead of pretending the call succeeded.
+		cmd.Println("Session extension call completed, but the management server did not return a new deadline (peer may not be SSO-tracked or login expiration is disabled).")
+	}
+	return nil
+}
+
 func getActiveProfile(ctx context.Context, pm *profilemanager.ProfileManager, profileName string, username string) (*profilemanager.Profile, error) {
 	// switch profile if provided
 

client/cmd/service.go

@@ -5,6 +5,7 @@ package cmd
 import (
 	"context"
 	"fmt"
+	"net/http"
 	"runtime"
 	"strings"
 	"sync"
@@ -22,15 +23,21 @@ var serviceCmd = &cobra.Command{
 	Short: "Manage the NetBird daemon service",
 }
 
+const defaultJSONSocket = "unix:///var/run/netbird-http.sock"
+
 var (
-	serviceName    string
-	serviceEnvVars []string
+	serviceName      string
+	serviceEnvVars   []string
+	jsonSocket       string
+	enableJSONSocket bool
 )
 
 type program struct {
 	ctx              context.Context
 	cancel           context.CancelFunc
 	serv             *grpc.Server
+	jsonServ         *http.Server
+	jsonServMu       sync.Mutex
 	serverInstance   *server.Server
 	serverInstanceMu sync.Mutex
 }
@@ -46,6 +53,8 @@ func init() {
 	serviceCmd.PersistentFlags().BoolVar(&updateSettingsDisabled, "disable-update-settings", false, "Disables update settings feature. If enabled, the client will not be able to change or edit any settings. To persist this setting, use: netbird service install --disable-update-settings")
 	serviceCmd.PersistentFlags().BoolVar(&captureEnabled, "enable-capture", false, "Enables packet capture via 'netbird debug capture'. To persist, use: netbird service install --enable-capture")
 	serviceCmd.PersistentFlags().BoolVar(&networksDisabled, "disable-networks", false, "Disables network selection. If enabled, the client will not allow listing, selecting, or deselecting networks. To persist, use: netbird service install --disable-networks")
+	serviceCmd.PersistentFlags().BoolVar(&enableJSONSocket, "enable-json-socket", false, "Enables the HTTP/JSON API socket served by grpc-gateway. To persist, use: netbird service install --enable-json-socket")
+	serviceCmd.PersistentFlags().StringVar(&jsonSocket, "json-socket", defaultJSONSocket, "HTTP/JSON API socket address [unix|tcp]://[path|host:port]. Requires --enable-json-socket to serve. To persist, use: netbird service install --enable-json-socket --json-socket")
 
 	rootCmd.PersistentFlags().StringVarP(&serviceName, "service", "s", defaultServiceName, "Netbird system service name")
 	serviceEnvDesc := `Sets extra environment variables for the service. ` +

client/cmd/service_controller.go

@@ -5,9 +5,6 @@ package cmd
 import (
 	"context"
 	"fmt"
-	"net"
-	"os"
-	"strings"
 	"time"
 
 	"github.com/kardianos/service"
@@ -22,41 +19,56 @@ import (
 	"github.com/netbirdio/netbird/util"
 )
 
+func validateJSONSocketFlags() error {
+	if serviceCmd.PersistentFlags().Changed("json-socket") && !enableJSONSocket {
+		return fmt.Errorf("--json-socket requires --enable-json-socket to configure the daemon JSON gateway")
+	}
+	return nil
+}
+
 func (p *program) Start(svc service.Service) error {
 	// Start should not block. Do the actual work async.
 	log.Info("starting NetBird service") //nolint
 
+	if err := validateJSONSocketFlags(); err != nil {
+		return err
+	}
+
 	// Collect static system and platform information
 	system.UpdateStaticInfoAsync()
 
 	// in any case, even if configuration does not exists we run daemon to serve CLI gRPC API.
 	p.serv = grpc.NewServer()
 
-	split := strings.Split(daemonAddr, "://")
-	switch split[0] {
-	case "unix":
-		// cleanup failed close
-		stat, err := os.Stat(split[1])
-		if err == nil && !stat.IsDir() {
-			if err := os.Remove(split[1]); err != nil {
-				log.Debugf("remove socket file: %v", err)
-			}
-		}
-	case "tcp":
-	default:
-		return fmt.Errorf("unsupported daemon address protocol: %v", split[0])
-	}
-
-	listen, err := net.Listen(split[0], split[1])
+	daemonListener, err := listenOnAddress(daemonAddr)
 	if err != nil {
 		return fmt.Errorf("listen daemon interface: %w", err)
 	}
-	go func() {
-		defer listen.Close()
 
-		if split[0] == "unix" {
-			if err := os.Chmod(split[1], 0666); err != nil {
-				log.Errorf("failed setting daemon permissions: %v", split[1])
+	var jsonListener *socketListener
+	if enableJSONSocket {
+		jsonListener, err = listenOnAddress(jsonSocket)
+		if err != nil {
+			_ = daemonListener.Close()
+			return fmt.Errorf("listen daemon JSON interface: %w", err)
+		}
+	} else {
+		removeStaleUnixSocketForAddress(jsonSocket)
+	}
+
+	go func() {
+		defer daemonListener.Close()
+		if jsonListener != nil {
+			defer jsonListener.Close()
+		}
+
+		if err := daemonListener.chmodUnixSocket("daemon"); err != nil {
+			log.Error(err)
+			return
+		}
+		if jsonListener != nil {
+			if err := jsonListener.chmodUnixSocket("daemon JSON"); err != nil {
+				log.Error(err)
 				return
 			}
 		}
@@ -71,8 +83,16 @@ func (p *program) Start(svc service.Service) error {
 		p.serverInstance = serverInstance
 		p.serverInstanceMu.Unlock()
 
-		log.Printf("started daemon server: %v", split[1])
-		if err := p.serv.Serve(listen); err != nil {
+		if jsonListener != nil {
+			if err := p.startJSONGateway(jsonListener, daemonAddr); err != nil {
+				log.Fatalf("failed to start daemon JSON server: %v", err)
+			}
+		} else {
+			log.Debug("daemon JSON socket disabled")
+		}
+
+		log.Printf("started daemon server: %v", daemonListener.address)
+		if err := p.serv.Serve(daemonListener.Listener); err != nil {
 			log.Errorf("failed to serve daemon requests: %v", err)
 		}
 	}()
@@ -92,6 +112,20 @@ func (p *program) Stop(srv service.Service) error {
 
 	p.cancel()
 
+	p.jsonServMu.Lock()
+	jsonServ := p.jsonServ
+	p.jsonServMu.Unlock()
+	if jsonServ != nil {
+		shutdownCtx, shutdownCancel := context.WithTimeout(context.Background(), 2*time.Second)
+		if err := jsonServ.Shutdown(shutdownCtx); err != nil {
+			log.Errorf("failed to stop daemon JSON server gracefully: %v", err)
+			if err := jsonServ.Close(); err != nil {
+				log.Errorf("failed to close daemon JSON server: %v", err)
+			}
+		}
+		shutdownCancel()
+	}
+
 	if p.serv != nil {
 		p.serv.Stop()
 	}
@@ -148,6 +182,9 @@ var runCmd = &cobra.Command{
 		if err != nil {
 			return err
 		}
+		if err := validateJSONSocketFlags(); err != nil {
+			return err
+		}
 
 		return s.Run()
 	},
@@ -162,6 +199,9 @@ var startCmd = &cobra.Command{
 		if err != nil {
 			return err
 		}
+		if err := validateJSONSocketFlags(); err != nil {
+			return err
+		}
 
 		if err := s.Start(); err != nil {
 			return fmt.Errorf("start service: %w", err)
@@ -198,6 +238,9 @@ var restartCmd = &cobra.Command{
 		if err != nil {
 			return err
 		}
+		if err := validateJSONSocketFlags(); err != nil {
+			return err
+		}
 
 		if err := s.Restart(); err != nil {
 			return fmt.Errorf("restart service: %w", err)

client/cmd/service_installer.go

@@ -67,6 +67,10 @@ func buildServiceArguments() []string {
 		args = append(args, "--disable-networks")
 	}
 
+	if enableJSONSocket {
+		args = append(args, "--enable-json-socket", "--json-socket", jsonSocket)
+	}
+
 	return args
 }
 
@@ -106,6 +110,10 @@ func configurePlatformSpecificSettings(svcConfig *service.Config) error {
 
 // Create fully configured service config for install/reconfigure
 func createServiceConfigForInstall() (*service.Config, error) {
+	if err := validateJSONSocketFlags(); err != nil {
+		return nil, err
+	}
+
 	svcConfig, err := newSVCConfig()
 	if err != nil {
 		return nil, fmt.Errorf("create service config: %w", err)

client/cmd/service_json_gateway.go

@@ -0,0 +1,52 @@
+//go:build !ios && !android
+
+package cmd
+
+import (
+	"context"
+	"errors"
+	"net"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/grpc-ecosystem/grpc-gateway/v2/runtime"
+	log "github.com/sirupsen/logrus"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/credentials/insecure"
+
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+func grpcGatewayEndpoint(addr string) string {
+	return strings.TrimPrefix(addr, "tcp://")
+}
+
+func (p *program) startJSONGateway(jsonListener *socketListener, daemonEndpoint string) error {
+	mux := runtime.NewServeMux()
+	opts := []grpc.DialOption{grpc.WithTransportCredentials(insecure.NewCredentials())}
+	if err := proto.RegisterDaemonServiceHandlerFromEndpoint(p.ctx, mux, grpcGatewayEndpoint(daemonEndpoint), opts); err != nil {
+		return err
+	}
+
+	jsonServer := &http.Server{
+		Handler:           mux,
+		ReadHeaderTimeout: 5 * time.Second,
+		BaseContext: func(net.Listener) context.Context {
+			return p.ctx
+		},
+	}
+
+	p.jsonServMu.Lock()
+	p.jsonServ = jsonServer
+	p.jsonServMu.Unlock()
+
+	go func() {
+		log.Printf("started daemon JSON server: %v", jsonListener.address)
+		if err := jsonServer.Serve(jsonListener.Listener); err != nil && !errors.Is(err, http.ErrServerClosed) {
+			log.Errorf("failed to serve daemon JSON requests: %v", err)
+		}
+	}()
+
+	return nil
+}

client/cmd/service_json_socket_test.go

@@ -0,0 +1,176 @@
+//go:build !ios && !android
+
+package cmd
+
+import (
+	"net"
+	"os"
+	"path/filepath"
+	"runtime"
+	"testing"
+
+	"github.com/spf13/cobra"
+	"github.com/spf13/pflag"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func preserveJSONSocketTestState(t *testing.T) {
+	t.Helper()
+
+	origJSONSocket := jsonSocket
+	origEnableJSONSocket := enableJSONSocket
+	origChanged := map[string]bool{}
+	serviceCmd.PersistentFlags().VisitAll(func(flag *pflag.Flag) {
+		origChanged[flag.Name] = flag.Changed
+	})
+
+	t.Cleanup(func() {
+		jsonSocket = origJSONSocket
+		enableJSONSocket = origEnableJSONSocket
+		serviceCmd.PersistentFlags().VisitAll(func(flag *pflag.Flag) {
+			flag.Changed = origChanged[flag.Name]
+		})
+	})
+}
+
+func TestJSONSocketFlagsArePositiveEnableOnly(t *testing.T) {
+	assert.NotNil(t, serviceCmd.PersistentFlags().Lookup("enable-json-socket"))
+	assert.NotNil(t, serviceCmd.PersistentFlags().Lookup("json-socket"))
+	assert.Nil(t, serviceCmd.PersistentFlags().Lookup("disable-json-socket"))
+	assert.Equal(t, "false", serviceCmd.PersistentFlags().Lookup("enable-json-socket").DefValue)
+}
+
+func TestBuildServiceArgumentsDefaultDisablesJSONSocket(t *testing.T) {
+	preserveJSONSocketTestState(t)
+
+	enableJSONSocket = false
+	jsonSocket = "tcp://127.0.0.1:8080"
+
+	args := buildServiceArguments()
+
+	assert.NotContains(t, args, "--enable-json-socket")
+	assert.NotContains(t, args, "--json-socket")
+}
+
+func TestBuildServiceArgumentsIncludesJSONSocketWhenEnabled(t *testing.T) {
+	preserveJSONSocketTestState(t)
+
+	enableJSONSocket = true
+	jsonSocket = "tcp://127.0.0.1:8080"
+
+	args := buildServiceArguments()
+
+	enableIndex := indexOfArg(args, "--enable-json-socket")
+	jsonIndex := indexOfArg(args, "--json-socket")
+	require.NotEqual(t, -1, enableIndex)
+	require.NotEqual(t, -1, jsonIndex)
+	require.Less(t, enableIndex, jsonIndex)
+	require.Less(t, jsonIndex+1, len(args))
+	assert.Equal(t, "tcp://127.0.0.1:8080", args[jsonIndex+1])
+}
+
+func TestJSONSocketWithoutEnableValidation(t *testing.T) {
+	preserveJSONSocketTestState(t)
+
+	enableJSONSocket = false
+	require.NoError(t, serviceCmd.PersistentFlags().Set("json-socket", "tcp://127.0.0.1:8080"))
+
+	err := validateJSONSocketFlags()
+
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "--enable-json-socket")
+}
+
+func TestJSONSocketWithEnableValidation(t *testing.T) {
+	preserveJSONSocketTestState(t)
+
+	require.NoError(t, serviceCmd.PersistentFlags().Set("enable-json-socket", "true"))
+	require.NoError(t, serviceCmd.PersistentFlags().Set("json-socket", "tcp://127.0.0.1:8080"))
+
+	assert.NoError(t, validateJSONSocketFlags())
+}
+
+func TestJSONSocketServiceParamsPersistEnableAndAddress(t *testing.T) {
+	preserveJSONSocketTestState(t)
+	serviceCmd.PersistentFlags().VisitAll(func(flag *pflag.Flag) {
+		flag.Changed = false
+	})
+
+	enableJSONSocket = true
+	jsonSocket = "tcp://127.0.0.1:8080"
+
+	params := currentServiceParams()
+	require.True(t, params.EnableJSONSocket)
+	require.Equal(t, "tcp://127.0.0.1:8080", params.JSONSocket)
+
+	enableJSONSocket = false
+	jsonSocket = defaultJSONSocket
+	applyServiceParams(testServiceEnvCommand(), params)
+
+	assert.True(t, enableJSONSocket)
+	assert.Equal(t, "tcp://127.0.0.1:8080", jsonSocket)
+}
+
+func TestRemoveStaleUnixSocketDoesNotRemoveRegularFile(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "netbird-http.sock")
+	require.NoError(t, os.WriteFile(path, []byte("not a socket"), 0600))
+
+	removeStaleUnixSocket(path)
+
+	data, err := os.ReadFile(path)
+	require.NoError(t, err)
+	assert.Equal(t, []byte("not a socket"), data)
+}
+
+func TestRemoveStaleUnixSocketRemovesSocket(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("unix sockets are not available on Windows")
+	}
+
+	path := filepath.Join(t.TempDir(), "netbird-http.sock")
+	addr := &net.UnixAddr{Name: path, Net: "unix"}
+	listener, err := net.ListenUnix("unix", addr)
+	require.NoError(t, err)
+	listener.SetUnlinkOnClose(false)
+	require.NoError(t, listener.Close())
+
+	_, err = os.Lstat(path)
+	require.NoError(t, err, "test setup must leave a stale Unix socket path")
+
+	removeStaleUnixSocket(path)
+
+	_, err = os.Lstat(path)
+	assert.True(t, os.IsNotExist(err), "expected stale Unix socket to be removed, got %v", err)
+}
+
+func TestRemoveStaleUnixSocketDoesNotRemoveLiveSocket(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("unix sockets are not available on Windows")
+	}
+
+	path := filepath.Join(t.TempDir(), "netbird-http.sock")
+	listener, err := net.Listen("unix", path)
+	require.NoError(t, err)
+	defer listener.Close()
+
+	removeStaleUnixSocket(path)
+
+	_, err = os.Lstat(path)
+	assert.NoError(t, err, "expected live Unix socket to be preserved")
+}
+
+func testServiceEnvCommand() *cobra.Command {
+	cmd := &cobra.Command{}
+	cmd.Flags().StringSlice("service-env", nil, "")
+	return cmd
+}
+
+func indexOfArg(args []string, arg string) int {
+	for i, candidate := range args {
+		if candidate == arg {
+			return i
+		}
+	}
+	return -1
+}

client/cmd/service_params.go

@@ -23,6 +23,7 @@ const serviceParamsFile = "service.json"
 type serviceParams struct {
 	LogLevel              string            `json:"log_level"`
 	DaemonAddr            string            `json:"daemon_addr"`
+	JSONSocket            string            `json:"json_socket"`
 	ManagementURL         string            `json:"management_url,omitempty"`
 	ConfigPath            string            `json:"config_path,omitempty"`
 	LogFiles              []string          `json:"log_files,omitempty"`
@@ -30,6 +31,7 @@ type serviceParams struct {
 	DisableUpdateSettings bool              `json:"disable_update_settings,omitempty"`
 	EnableCapture         bool              `json:"enable_capture,omitempty"`
 	DisableNetworks       bool              `json:"disable_networks,omitempty"`
+	EnableJSONSocket      bool              `json:"enable_json_socket,omitempty"`
 	ServiceEnvVars        map[string]string `json:"service_env_vars,omitempty"`
 }
 
@@ -75,6 +77,7 @@ func currentServiceParams() *serviceParams {
 	params := &serviceParams{
 		LogLevel:              logLevel,
 		DaemonAddr:            daemonAddr,
+		JSONSocket:            jsonSocket,
 		ManagementURL:         managementURL,
 		ConfigPath:            configPath,
 		LogFiles:              logFiles,
@@ -82,6 +85,7 @@ func currentServiceParams() *serviceParams {
 		DisableUpdateSettings: updateSettingsDisabled,
 		EnableCapture:         captureEnabled,
 		DisableNetworks:       networksDisabled,
+		EnableJSONSocket:      enableJSONSocket,
 	}
 
 	if len(serviceEnvVars) > 0 {
@@ -113,9 +117,8 @@ func applyServiceParams(cmd *cobra.Command, params *serviceParams) {
 		return
 	}
 
-	// For fields with non-empty defaults (log-level, daemon-addr), keep the
-	// != "" guard so that an older service.json missing the field doesn't
-	// clobber the default with an empty string.
+	// For fields with non-empty defaults, keep the != "" guard so that an older
+	// service.json missing the field doesn't clobber the default with an empty string.
 	if !rootCmd.PersistentFlags().Changed("log-level") && params.LogLevel != "" {
 		logLevel = params.LogLevel
 	}
@@ -124,6 +127,14 @@ func applyServiceParams(cmd *cobra.Command, params *serviceParams) {
 		daemonAddr = params.DaemonAddr
 	}
 
+	if !serviceCmd.PersistentFlags().Changed("json-socket") && params.JSONSocket != "" {
+		jsonSocket = params.JSONSocket
+	}
+
+	if !serviceCmd.PersistentFlags().Changed("enable-json-socket") {
+		enableJSONSocket = params.EnableJSONSocket
+	}
+
 	// For optional fields where empty means "use default", always apply so
 	// that an explicit clear (--management-url "") persists across reinstalls.
 	if !rootCmd.PersistentFlags().Changed("management-url") {

client/cmd/service_params_test.go

@@ -41,6 +41,8 @@ func TestSaveAndLoadServiceParams(t *testing.T) {
 	params := &serviceParams{
 		LogLevel:              "debug",
 		DaemonAddr:            "unix:///var/run/netbird.sock",
+		JSONSocket:            "tcp://127.0.0.1:8080",
+		EnableJSONSocket:      true,
 		ManagementURL:         "https://my.server.com",
 		ConfigPath:            "/etc/netbird/config.json",
 		LogFiles:              []string{"/var/log/netbird/client.log", "console"},
@@ -63,6 +65,8 @@ func TestSaveAndLoadServiceParams(t *testing.T) {
 
 	assert.Equal(t, params.LogLevel, loaded.LogLevel)
 	assert.Equal(t, params.DaemonAddr, loaded.DaemonAddr)
+	assert.Equal(t, params.JSONSocket, loaded.JSONSocket)
+	assert.Equal(t, params.EnableJSONSocket, loaded.EnableJSONSocket)
 	assert.Equal(t, params.ManagementURL, loaded.ManagementURL)
 	assert.Equal(t, params.ConfigPath, loaded.ConfigPath)
 	assert.Equal(t, params.LogFiles, loaded.LogFiles)
@@ -101,6 +105,8 @@ func TestLoadServiceParams_InvalidJSON(t *testing.T) {
 func TestCurrentServiceParams(t *testing.T) {
 	origLogLevel := logLevel
 	origDaemonAddr := daemonAddr
+	origJSONSocket := jsonSocket
+	origEnableJSONSocket := enableJSONSocket
 	origManagementURL := managementURL
 	origConfigPath := configPath
 	origLogFiles := logFiles
@@ -110,6 +116,8 @@ func TestCurrentServiceParams(t *testing.T) {
 	t.Cleanup(func() {
 		logLevel = origLogLevel
 		daemonAddr = origDaemonAddr
+		jsonSocket = origJSONSocket
+		enableJSONSocket = origEnableJSONSocket
 		managementURL = origManagementURL
 		configPath = origConfigPath
 		logFiles = origLogFiles
@@ -120,6 +128,8 @@ func TestCurrentServiceParams(t *testing.T) {
 
 	logLevel = "trace"
 	daemonAddr = "tcp://127.0.0.1:9999"
+	jsonSocket = "tcp://127.0.0.1:8080"
+	enableJSONSocket = true
 	managementURL = "https://mgmt.example.com"
 	configPath = "/tmp/test-config.json"
 	logFiles = []string{"/tmp/test.log"}
@@ -131,6 +141,8 @@ func TestCurrentServiceParams(t *testing.T) {
 
 	assert.Equal(t, "trace", params.LogLevel)
 	assert.Equal(t, "tcp://127.0.0.1:9999", params.DaemonAddr)
+	assert.Equal(t, "tcp://127.0.0.1:8080", params.JSONSocket)
+	assert.True(t, params.EnableJSONSocket)
 	assert.Equal(t, "https://mgmt.example.com", params.ManagementURL)
 	assert.Equal(t, "/tmp/test-config.json", params.ConfigPath)
 	assert.Equal(t, []string{"/tmp/test.log"}, params.LogFiles)
@@ -142,6 +154,8 @@ func TestCurrentServiceParams(t *testing.T) {
 func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 	origLogLevel := logLevel
 	origDaemonAddr := daemonAddr
+	origJSONSocket := jsonSocket
+	origEnableJSONSocket := enableJSONSocket
 	origManagementURL := managementURL
 	origConfigPath := configPath
 	origLogFiles := logFiles
@@ -151,6 +165,8 @@ func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 	t.Cleanup(func() {
 		logLevel = origLogLevel
 		daemonAddr = origDaemonAddr
+		jsonSocket = origJSONSocket
+		enableJSONSocket = origEnableJSONSocket
 		managementURL = origManagementURL
 		configPath = origConfigPath
 		logFiles = origLogFiles
@@ -162,6 +178,8 @@ func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 	// Reset all flags to defaults.
 	logLevel = "info"
 	daemonAddr = "unix:///var/run/netbird.sock"
+	jsonSocket = defaultJSONSocket
+	enableJSONSocket = false
 	managementURL = ""
 	configPath = "/etc/netbird/config.json"
 	logFiles = []string{"/var/log/netbird/client.log"}
@@ -184,6 +202,8 @@ func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 	saved := &serviceParams{
 		LogLevel:              "debug",
 		DaemonAddr:            "tcp://127.0.0.1:5555",
+		JSONSocket:            "tcp://127.0.0.1:8080",
+		EnableJSONSocket:      true,
 		ManagementURL:         "https://saved.example.com",
 		ConfigPath:            "/saved/config.json",
 		LogFiles:              []string{"/saved/client.log"},
@@ -201,6 +221,8 @@ func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 
 	// All other fields were not Changed, so they should use saved values.
 	assert.Equal(t, "tcp://127.0.0.1:5555", daemonAddr)
+	assert.Equal(t, "tcp://127.0.0.1:8080", jsonSocket)
+	assert.True(t, enableJSONSocket)
 	assert.Equal(t, "https://saved.example.com", managementURL)
 	assert.Equal(t, "/saved/config.json", configPath)
 	assert.Equal(t, []string{"/saved/client.log"}, logFiles)
@@ -212,14 +234,17 @@ func TestApplyServiceParams_OnlyUnchangedFlags(t *testing.T) {
 func TestApplyServiceParams_BooleanRevertToFalse(t *testing.T) {
 	origProfilesDisabled := profilesDisabled
 	origUpdateSettingsDisabled := updateSettingsDisabled
+	origEnableJSONSocket := enableJSONSocket
 	t.Cleanup(func() {
 		profilesDisabled = origProfilesDisabled
 		updateSettingsDisabled = origUpdateSettingsDisabled
+		enableJSONSocket = origEnableJSONSocket
 	})
 
 	// Simulate current state where booleans are true (e.g. set by previous install).
 	profilesDisabled = true
 	updateSettingsDisabled = true
+	enableJSONSocket = true
 
 	// Reset Changed state so flags appear unset.
 	serviceCmd.PersistentFlags().VisitAll(func(f *pflag.Flag) {
@@ -238,6 +263,7 @@ func TestApplyServiceParams_BooleanRevertToFalse(t *testing.T) {
 
 	assert.False(t, profilesDisabled, "saved false should override current true")
 	assert.False(t, updateSettingsDisabled, "saved false should override current true")
+	assert.False(t, enableJSONSocket, "saved false should override current true")
 }
 
 func TestApplyServiceParams_ClearManagementURL(t *testing.T) {
@@ -530,6 +556,7 @@ func fieldToGlobalVar(field string) string {
 	m := map[string]string{
 		"LogLevel":              "logLevel",
 		"DaemonAddr":            "daemonAddr",
+		"JSONSocket":            "jsonSocket",
 		"ManagementURL":         "managementURL",
 		"ConfigPath":            "configPath",
 		"LogFiles":              "logFiles",
@@ -537,6 +564,7 @@ func fieldToGlobalVar(field string) string {
 		"DisableUpdateSettings": "updateSettingsDisabled",
 		"EnableCapture":         "captureEnabled",
 		"DisableNetworks":       "networksDisabled",
+		"EnableJSONSocket":      "enableJSONSocket",
 		"ServiceEnvVars":        "serviceEnvVars",
 	}
 	if v, ok := m[field]; ok {

client/cmd/service_socket.go

@@ -0,0 +1,111 @@
+//go:build !ios && !android
+
+package cmd
+
+import (
+	"errors"
+	"fmt"
+	"net"
+	"os"
+	"strings"
+	"syscall"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+)
+
+type socketListener struct {
+	net.Listener
+	network string
+	address string
+}
+
+func listenOnAddress(addr string) (*socketListener, error) {
+	network, address, err := parseListenAddress(addr)
+	if err != nil {
+		return nil, err
+	}
+
+	if network == "unix" {
+		removeStaleUnixSocket(address)
+	}
+
+	listener, err := net.Listen(network, address)
+	if err != nil {
+		return nil, err
+	}
+
+	return &socketListener{Listener: listener, network: network, address: address}, nil
+}
+
+func parseListenAddress(addr string) (string, string, error) {
+	network, address, ok := strings.Cut(addr, "://")
+	if !ok || network == "" || address == "" {
+		return "", "", fmt.Errorf("address must be in [unix|tcp]://[path|host:port] format: %q", addr)
+	}
+
+	switch network {
+	case "unix", "tcp":
+		return network, address, nil
+	default:
+		return "", "", fmt.Errorf("unsupported daemon address protocol: %v", network)
+	}
+}
+
+func removeStaleUnixSocket(path string) {
+	stat, err := os.Lstat(path)
+	if err != nil {
+		if !os.IsNotExist(err) {
+			log.Debugf("stat socket file: %v", err)
+		}
+		return
+	}
+
+	if stat.Mode()&os.ModeSocket == 0 {
+		return
+	}
+
+	if !isStaleUnixSocket(path) {
+		return
+	}
+
+	if err := os.Remove(path); err != nil {
+		log.Debugf("remove socket file: %v", err)
+	}
+}
+
+func isStaleUnixSocket(path string) bool {
+	conn, err := net.DialTimeout("unix", path, 100*time.Millisecond)
+	if err == nil {
+		if closeErr := conn.Close(); closeErr != nil {
+			log.Debugf("close unix socket probe: %v", closeErr)
+		}
+		return false
+	}
+
+	if os.IsNotExist(err) || os.IsPermission(err) || os.IsTimeout(err) {
+		log.Debugf("not removing unix socket %s after probe error: %v", path, err)
+		return false
+	}
+
+	return errors.Is(err, syscall.ECONNREFUSED)
+}
+
+func removeStaleUnixSocketForAddress(addr string) {
+	network, address, err := parseListenAddress(addr)
+	if err != nil || network != "unix" {
+		return
+	}
+	removeStaleUnixSocket(address)
+}
+
+func (l *socketListener) chmodUnixSocket(description string) error {
+	if l == nil || l.network != "unix" {
+		return nil
+	}
+
+	if err := os.Chmod(l.address, 0666); err != nil {
+		return fmt.Errorf("failed setting %s permissions for %s: %w", description, l.address, err)
+	}
+	return nil
+}

client/cmd/status.go

@@ -6,6 +6,7 @@ import (
 	"net"
 	"net/netip"
 	"strings"
+	"time"
 
 	"github.com/spf13/cobra"
 	"google.golang.org/grpc/status"
@@ -115,6 +116,11 @@ func statusFunc(cmd *cobra.Command, args []string) error {
 	// manager only knows the active profile ID, not its display name.
 	profName := getActiveProfileName(ctx)
 
+	var sessionExpiresAt time.Time
+	if ts := resp.GetSessionExpiresAt(); ts.IsValid() {
+		sessionExpiresAt = ts.AsTime().UTC()
+	}
+
 	var outputInformationHolder = nbstatus.ConvertToStatusOutputOverview(resp.GetFullStatus(), nbstatus.ConvertOptions{
 		Anonymize:            anonymizeFlag,
 		DaemonVersion:        resp.GetDaemonVersion(),
@@ -125,6 +131,7 @@ func statusFunc(cmd *cobra.Command, args []string) error {
 		IPsFilter:            ipsFilterMap,
 		ConnectionTypeFilter: connectionTypeFilter,
 		ProfileName:          profName,
+		SessionExpiresAt:     sessionExpiresAt,
 	})
 	var statusOutputString string
 	switch {

client/embed/embed.go

@@ -470,7 +470,7 @@ func (c *Client) Status() (peer.FullStatus, error) {
 	if connect != nil {
 		engine := connect.Engine()
 		if engine != nil {
-			_ = engine.RunHealthProbes(false)
+			_ = engine.RunHealthProbes(context.Background(), false)
 		}
 	}
 

client/installer.nsis

@@ -6,7 +6,7 @@
 !define DESCRIPTION "Connect your devices into a secure WireGuard-based overlay network with SSO, MFA, and granular access controls."
 !define INSTALLER_NAME "netbird-installer.exe"
 !define MAIN_APP_EXE "Netbird"
-!define ICON "ui\\assets\\netbird.ico"
+!define ICON "ui\\build\\windows\\icon.ico"
 !define BANNER "ui\\build\\banner.bmp"
 !define LICENSE_DATA "..\\LICENSE"
 
@@ -79,8 +79,6 @@ ShowInstDetails Show
 
 !insertmacro MUI_PAGE_DIRECTORY
 
-Page custom AutostartPage AutostartPageLeave
-
 !insertmacro MUI_PAGE_INSTFILES
 
 !insertmacro MUI_PAGE_FINISH
@@ -97,40 +95,12 @@ UninstPage custom un.DeleteDataPage un.DeleteDataPageLeave
 
 !insertmacro MUI_LANGUAGE "English"
 
-; Variables for autostart option
-Var AutostartCheckbox
-Var AutostartEnabled
-
 ; Variables for uninstall data deletion option
 Var DeleteDataCheckbox
 Var DeleteDataEnabled
 
 ######################################################################
 
-; Function to create the autostart options page
-Function AutostartPage
-  !insertmacro MUI_HEADER_TEXT "Startup Options" "Configure how ${APP_NAME} launches with Windows."
-
-  nsDialogs::Create 1018
-  Pop $0
-
-  ${If} $0 == error
-    Abort
-  ${EndIf}
-
-  ${NSD_CreateCheckbox} 0 20u 100% 10u "Start ${APP_NAME} UI automatically when Windows starts"
-  Pop $AutostartCheckbox
-  ${NSD_Check} $AutostartCheckbox
-  StrCpy $AutostartEnabled "1"
-
-  nsDialogs::Show
-FunctionEnd
-
-; Function to handle leaving the autostart page
-Function AutostartPageLeave
-  ${NSD_GetState} $AutostartCheckbox $AutostartEnabled
-FunctionEnd
-
 ; Function to create the uninstall data deletion page
 Function un.DeleteDataPage
   !insertmacro MUI_HEADER_TEXT "Uninstall Options" "Choose whether to delete ${APP_NAME} data."
@@ -201,8 +171,6 @@ Pop $0
 
 Function .onInit
 StrCpy $INSTDIR "${INSTALL_DIR}"
-; Default autostart to enabled so silent installs (/S) match the interactive default
-StrCpy $AutostartEnabled "1"
 
 ; Pre-0.70.1 installers ran without SetRegView, so their uninstall keys live
 ; in the 32-bit view. Fall back to it so upgrades still find them.
@@ -260,17 +228,12 @@ WriteRegStr ${REG_ROOT} "${UNINSTALL_PATH}"  "Publisher" "${COMP_NAME}"
 
 WriteRegStr ${REG_ROOT} "${UI_REG_APP_PATH}" "" "$INSTDIR\${UI_APP_EXE}"
 
-; Create autostart registry entry based on checkbox
-DetailPrint "Autostart enabled: $AutostartEnabled"
-${If} $AutostartEnabled == "1"
-    WriteRegStr HKLM "${AUTOSTART_REG_KEY}" "${APP_NAME}" '"$INSTDIR\${UI_APP_EXE}.exe"'
-    DetailPrint "Added autostart registry entry: $INSTDIR\${UI_APP_EXE}.exe"
-${Else}
-    DeleteRegValue HKLM "${AUTOSTART_REG_KEY}" "${APP_NAME}"
-    ; Legacy: pre-HKLM installs wrote to HKCU; clean that up too.
-    DeleteRegValue HKCU "${AUTOSTART_REG_KEY}" "${APP_NAME}"
-    DetailPrint "Autostart not enabled by user"
-${EndIf}
+; Autostart is owned by the UI's per-user setting (HKCU\...\Run via Wails),
+; not the installer. Drop the machine-wide entry older installers wrote so the
+; toggle is the single source of truth. HKCU is left untouched -- it may hold
+; the user's own toggle state, which must survive upgrades.
+DetailPrint "Removing installer-managed autostart registry entry if present..."
+DeleteRegValue HKLM "${AUTOSTART_REG_KEY}" "${APP_NAME}"
 
 EnVar::SetHKLM
 EnVar::AddValueEx "path" "$INSTDIR"
@@ -280,6 +243,43 @@ CreateShortCut "$SMPROGRAMS\${APP_NAME}.lnk" "$INSTDIR\${UI_APP_EXE}"
 CreateShortCut "$DESKTOP\${APP_NAME}.lnk" "$INSTDIR\${UI_APP_EXE}"
 SectionEnd
 
+# Install the Microsoft Edge WebView2 runtime if it isn't already present.
+# Macro adapted from Wails3's NSIS template (wails_tools.nsh): a registry
+# probe followed by a silent install of the embedded evergreen bootstrapper.
+# The MicrosoftEdgeWebview2Setup.exe payload is staged next to this script
+# by the sign-pipelines build step (`wails3 generate webview2bootstrapper`).
+!macro nb.webview2runtime
+    SetRegView 64
+    # Per-machine install marker — populated when the runtime ships with
+    # Edge or has been installed by an admin previously.
+    ReadRegStr $0 HKLM "SOFTWARE\WOW6432Node\Microsoft\EdgeUpdate\Clients\{F3017226-FE2A-4295-8BDF-00C3A9A7E4C5}" "pv"
+    ${If} $0 != ""
+        Goto webview2_ok
+    ${EndIf}
+    # Per-user fallback for HKCU installs.
+    ReadRegStr $0 HKCU "Software\Microsoft\EdgeUpdate\Clients\{F3017226-FE2A-4295-8BDF-00C3A9A7E4C5}" "pv"
+    ${If} $0 != ""
+        Goto webview2_ok
+    ${EndIf}
+
+    SetDetailsPrint both
+    DetailPrint "Installing: WebView2 Runtime"
+    SetDetailsPrint listonly
+
+    InitPluginsDir
+    CreateDirectory "$pluginsdir\webview2bootstrapper"
+    SetOutPath "$pluginsdir\webview2bootstrapper"
+    File "MicrosoftEdgeWebview2Setup.exe"
+    ExecWait '"$pluginsdir\webview2bootstrapper\MicrosoftEdgeWebview2Setup.exe" /silent /install'
+
+    SetDetailsPrint both
+    webview2_ok:
+!macroend
+
+Section -WebView2
+    !insertmacro nb.webview2runtime
+SectionEnd
+
 Section -Post
 ExecWait '"$INSTDIR\${MAIN_APP_EXE}" service install'
 ExecWait '"$INSTDIR\${MAIN_APP_EXE}" service start'
@@ -299,11 +299,14 @@ ExecWait '"$INSTDIR\${MAIN_APP_EXE}" service uninstall'
 DetailPrint "Terminating Netbird UI process..."
 ExecWait `taskkill /im ${UI_APP_EXE}.exe /f`
 
-; Remove autostart registry entry
-DetailPrint "Removing autostart registry entry if exists..."
+; Remove autostart registry entries
+DetailPrint "Removing autostart registry entries if they exist..."
+; Legacy machine-wide entry written by older installers.
 DeleteRegValue HKLM "${AUTOSTART_REG_KEY}" "${APP_NAME}"
-; Legacy: pre-HKLM installs wrote to HKCU; clean that up too.
+; Per-user entry the UI toggle writes via Wails (value name is the lowercase
+; app-name slug). Uninstall removes the app, so drop it too.
 DeleteRegValue HKCU "${AUTOSTART_REG_KEY}" "${APP_NAME}"
+DeleteRegValue HKCU "${AUTOSTART_REG_KEY}" "netbird"
 
 ; Handle data deletion based on checkbox
 DetailPrint "Checking if user requested data deletion..."
@@ -326,9 +329,9 @@ DetailPrint "Deleting application files..."
 Delete "$INSTDIR\${UI_APP_EXE}"
 Delete "$INSTDIR\${MAIN_APP_EXE}"
 Delete "$INSTDIR\wintun.dll"
-!if ${ARCH} == "amd64"
+# Legacy: pre-Wails installs shipped opengl32.dll (Mesa3D for Fyne); remove
+# any leftover copy on uninstall so old upgrades don't leave it behind.
 Delete "$INSTDIR\opengl32.dll"
-!endif
 DetailPrint "Removing application directory..."
 RmDir /r "$INSTDIR"
 

client/internal/auth/auth.go

@@ -3,6 +3,7 @@ package auth
 import (
 	"context"
 	"net/url"
+	"strings"
 	"sync"
 	"time"
 
@@ -21,6 +22,25 @@ import (
 	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
+// peerLoginExpiredMsg is the exact phrase the management server returns
+// when a previously SSO-enrolled peer's login has expired. Sourced from
+// shared/management/status/error.go (NewPeerLoginExpiredError). Matched
+// by substring so a future server-side rewording that keeps the phrase
+// still triggers the friendly fallback in Login().
+const peerLoginExpiredMsg = "peer login has expired"
+
+// errSetupKeyOnSSOExpiredPeer replaces the raw management error when the
+// user runs `netbird login -k <setup-key>` against a peer that was
+// originally enrolled via SSO. Wrapped in a PermissionDenied gRPC status
+// so callers' existing isPermissionDenied / isAuthError checks still
+// classify it correctly (early-exit from retry backoff, StatusNeedsLogin
+// in the server state machine).
+var errSetupKeyOnSSOExpiredPeer = status.Error(
+	codes.PermissionDenied,
+	"this peer was originally enrolled via SSO and its session has expired. "+
+		"Setup keys can only enrol new peers — run `netbird up` (interactive SSO) to re-login.",
+)
+
 // Auth manages authentication operations with the management server
 // It maintains a long-lived connection and automatically handles reconnection with backoff
 type Auth struct {
@@ -184,6 +204,15 @@ func (a *Auth) Login(ctx context.Context, setupKey string, jwtToken string) (err
 			log.Debugf("peer registration required")
 			_, err = a.registerPeer(client, ctx, setupKey, jwtToken, pubSSHKey)
 			if err != nil {
+				// The peer pub-key is already on file with the management
+				// server (originally enrolled via SSO) and the session has
+				// expired. The setup-key path can only enrol new peers, so
+				// retrying with -k will keep failing. Replace the raw mgm
+				// message with an actionable hint that tells the user to
+				// re-authenticate via SSO instead.
+				if setupKey != "" && jwtToken == "" && isPeerLoginExpired(err) {
+					err = errSetupKeyOnSSOExpiredPeer
+				}
 				isAuthError = isPermissionDenied(err)
 				return err
 			}
@@ -473,3 +502,16 @@ func isLoginNeeded(err error) bool {
 func isRegistrationNeeded(err error) bool {
 	return isPermissionDenied(err)
 }
+
+// isPeerLoginExpired reports whether err is the management server's
+// "peer login has expired" PermissionDenied response. Used by Login to
+// detect the case where the caller passed a setup-key but the peer is
+// actually an SSO-enrolled record whose session needs refreshing — the
+// setup-key path cannot help there.
+func isPeerLoginExpired(err error) bool {
+	if !isPermissionDenied(err) {
+		return false
+	}
+	s, _ := status.FromError(err)
+	return strings.Contains(s.Message(), peerLoginExpiredMsg)
+}

client/internal/auth/auth_test.go

@@ -0,0 +1,80 @@
+package auth
+
+import (
+	"errors"
+	"strings"
+	"testing"
+
+	"google.golang.org/grpc/codes"
+	"google.golang.org/grpc/status"
+)
+
+func TestIsPeerLoginExpired(t *testing.T) {
+	cases := []struct {
+		name string
+		err  error
+		want bool
+	}{
+		{
+			name: "nil",
+			err:  nil,
+			want: false,
+		},
+		{
+			name: "plain error (not a gRPC status)",
+			err:  errors.New("network read: connection reset"),
+			want: false,
+		},
+		{
+			name: "PermissionDenied with different message",
+			err:  status.Error(codes.PermissionDenied, "user is blocked"),
+			want: false,
+		},
+		{
+			name: "Unauthenticated with the expected phrase",
+			// Wrong status code — must still return false.
+			err:  status.Error(codes.Unauthenticated, "peer login has expired, please log in once more"),
+			want: false,
+		},
+		{
+			name: "exact server message",
+			err:  status.Error(codes.PermissionDenied, "peer login has expired, please log in once more"),
+			want: true,
+		},
+		{
+			name: "phrase as substring",
+			// Future-proofing: if mgm reworords but keeps the phrase,
+			// the friendly fallback must still kick in.
+			err:  status.Error(codes.PermissionDenied, "session refused: peer login has expired (account=foo)"),
+			want: true,
+		},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			if got := isPeerLoginExpired(tc.err); got != tc.want {
+				t.Fatalf("isPeerLoginExpired(%v) = %v, want %v", tc.err, got, tc.want)
+			}
+		})
+	}
+}
+
+func TestErrSetupKeyOnSSOExpiredPeer(t *testing.T) {
+	// Sentinel must surface as PermissionDenied so the upstream
+	// isPermissionDenied / isAuthError checks classify it correctly
+	// (short-circuit retry backoff, set StatusNeedsLogin).
+	if !isPermissionDenied(errSetupKeyOnSSOExpiredPeer) {
+		t.Fatalf("errSetupKeyOnSSOExpiredPeer must be a PermissionDenied gRPC error")
+	}
+
+	// Message must actually mention SSO and `netbird up` so it is
+	// actionable for the end user. Loose substring checks keep the
+	// test resilient to copy edits.
+	s, _ := status.FromError(errSetupKeyOnSSOExpiredPeer)
+	msg := strings.ToLower(s.Message())
+	for _, want := range []string{"sso", "netbird up"} {
+		if !strings.Contains(msg, want) {
+			t.Errorf("sentinel message should contain %q, got %q", want, s.Message())
+		}
+	}
+}

client/internal/auth/device_flow.go

@@ -259,12 +259,18 @@ func (d *DeviceAuthorizationFlow) WaitToken(ctx context.Context, info AuthFlowIn
 	ticker := time.NewTicker(interval)
 	defer ticker.Stop()
 
+	log.Infof("device flow: waiting for user authorization, polling token endpoint every %s, code expires in %s", interval, timeout)
+
+	start := time.Now()
+	polls := 0
+
 	for {
 		select {
 		case <-waitCtx.Done():
 			return TokenInfo{}, waitCtx.Err()
 		case <-ticker.C:
 
+			polls++
 			tokenResponse, err := d.requestToken(info)
 			if err != nil {
 				return TokenInfo{}, fmt.Errorf("parsing token response failed with error: %v", err)
@@ -272,10 +278,12 @@ func (d *DeviceAuthorizationFlow) WaitToken(ctx context.Context, info AuthFlowIn
 
 			if tokenResponse.Error != "" {
 				if tokenResponse.Error == "authorization_pending" {
+					log.Tracef("device flow: authorization still pending after poll %d", polls)
 					continue
 				} else if tokenResponse.Error == "slow_down" {
 					interval += (3 * time.Second)
 					ticker.Reset(interval)
+					log.Infof("device flow: IdP requested slow_down, polling interval increased to %s", interval)
 					continue
 				}
 
@@ -296,6 +304,7 @@ func (d *DeviceAuthorizationFlow) WaitToken(ctx context.Context, info AuthFlowIn
 				return TokenInfo{}, fmt.Errorf("validate access token failed with error: %v", err)
 			}
 
+			log.Infof("device flow: user authorization confirmed after %d polls in %s", polls, time.Since(start).Round(time.Second))
 			return tokenInfo, err
 		}
 	}

client/internal/auth/pending_flow.go

@@ -0,0 +1,89 @@
+package auth
+
+import (
+	"context"
+	"sync"
+	"time"
+)
+
+// PendingFlow stores an in-progress OAuth flow between the RPC that
+// initiates it (returns the verification URI to the UI) and the RPC
+// that waits for the user to complete it. The flow handle, the
+// device-code info, and the absolute expiry are kept together so the
+// waiting RPC can validate the device code and reuse the same flow.
+//
+// PendingFlow is safe for concurrent use; callers must not access the
+// stored fields directly.
+type PendingFlow struct {
+	mu         sync.Mutex
+	flow       OAuthFlow
+	info       AuthFlowInfo
+	expiresAt  time.Time
+	waitCancel context.CancelFunc
+}
+
+// NewPendingFlow returns an empty PendingFlow ready to be populated by Set.
+func NewPendingFlow() *PendingFlow {
+	return &PendingFlow{}
+}
+
+// Set stores the flow and its authorization info, computing the absolute
+// expiry from info.ExpiresIn (seconds, as returned by the IdP).
+func (p *PendingFlow) Set(flow OAuthFlow, info AuthFlowInfo) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.flow = flow
+	p.info = info
+	p.expiresAt = time.Now().Add(time.Duration(info.ExpiresIn) * time.Second)
+}
+
+// Get returns the stored flow, info, and whether a flow is currently
+// pending. Returns (nil, zero, false) after Clear or before Set.
+func (p *PendingFlow) Get() (OAuthFlow, AuthFlowInfo, bool) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	if p.flow == nil {
+		return nil, AuthFlowInfo{}, false
+	}
+	return p.flow, p.info, true
+}
+
+// ExpiresAt returns the absolute expiry of the pending flow. Returns
+// the zero time when no flow is pending.
+func (p *PendingFlow) ExpiresAt() time.Time {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	return p.expiresAt
+}
+
+// SetWaitCancel records the cancel function for the goroutine currently
+// blocked in WaitToken so a new RequestAuth can preempt it.
+func (p *PendingFlow) SetWaitCancel(cancel context.CancelFunc) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.waitCancel = cancel
+}
+
+// CancelWait invokes and clears the stored wait-cancel, if any. Safe to
+// call when no wait is in progress.
+func (p *PendingFlow) CancelWait() {
+	p.mu.Lock()
+	cancel := p.waitCancel
+	p.waitCancel = nil
+	p.mu.Unlock()
+	if cancel != nil {
+		cancel()
+	}
+}
+
+// Clear resets the pending flow to empty. Any stored wait-cancel is
+// dropped without being invoked — call CancelWait first if the waiting
+// goroutine must be stopped.
+func (p *PendingFlow) Clear() {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+	p.flow = nil
+	p.info = AuthFlowInfo{}
+	p.expiresAt = time.Time{}
+	p.waitCancel = nil
+}

client/internal/auth/pkce_flow.go

@@ -188,6 +188,8 @@ func (p *PKCEAuthorizationFlow) WaitToken(ctx context.Context, info AuthFlowInfo
 	waitCtx, cancel := context.WithTimeout(ctx, timeout)
 	defer cancel()
 
+	log.Infof("pkce flow: waiting for authorization callback on %s, timeout %s", p.oAuthConfig.RedirectURL, timeout)
+
 	tokenChan := make(chan *oauth2.Token, 1)
 	errChan := make(chan error, 1)
 
@@ -221,6 +223,7 @@ func (p *PKCEAuthorizationFlow) WaitToken(ctx context.Context, info AuthFlowInfo
 func (p *PKCEAuthorizationFlow) startServer(server *http.Server, tokenChan chan<- *oauth2.Token, errChan chan<- error) {
 	mux := http.NewServeMux()
 	mux.HandleFunc("/", func(w http.ResponseWriter, req *http.Request) {
+		log.Infof("pkce flow: received authorization callback from IdP")
 		cert := p.providerConfig.ClientCertPair
 		if cert != nil {
 			tr := &http.Transport{
@@ -271,11 +274,18 @@ func (p *PKCEAuthorizationFlow) handleRequest(req *http.Request) (*oauth2.Token,
 		return nil, fmt.Errorf("authentication failed: missing code")
 	}
 
-	return p.oAuthConfig.Exchange(
+	exchangeStart := time.Now()
+	token, err := p.oAuthConfig.Exchange(
 		req.Context(),
 		code,
 		oauth2.SetAuthURLParam("code_verifier", p.codeVerifier),
 	)
+	if err != nil {
+		return nil, err
+	}
+
+	log.Infof("pkce flow: authorization code exchanged for token in %s", time.Since(exchangeStart).Round(time.Millisecond))
+	return token, nil
 }
 
 func (p *PKCEAuthorizationFlow) parseOAuthToken(token *oauth2.Token) (TokenInfo, error) {

client/internal/auth/sessionwatch/event.go

@@ -0,0 +1,82 @@
+package sessionwatch
+
+import (
+	"strconv"
+	"time"
+)
+
+// internal event kinds are no longer exposed: the watcher drives the Sink
+// directly (NotifyStateChange on deadline change/clear, PublishEvent at
+// each warning lead). Tests use a mock Sink to observe what the watcher
+// emits.
+
+// Metadata keys attached by the daemon to session-warning SystemEvents.
+// The UI tray reads these to build a locale-aware notification without
+// relying on the daemon's locale-less UserMessage string, and to
+// disambiguate the T-WarningLead notification from the T-FinalWarningLead
+// fallback that auto-opens the SessionAboutToExpire dialog.
+const (
+	// MetaSessionWarning is set to "true" on both warning events (T-10 and
+	// T-2) so the UI can detect a session-warning SystemEvent without
+	// matching on the message text. Use MetaSessionFinal to distinguish
+	// the two.
+	MetaSessionWarning = "session_warning"
+	// MetaSessionFinal is set to "true" on the T-FinalWarningLead event
+	// only. Consumers that need to auto-open the SessionAboutToExpire
+	// dialog gate on this; T-WarningLead events leave the field unset.
+	MetaSessionFinal = "session_final_warning"
+	// MetaSessionExpiresAt carries the absolute UTC deadline encoded with
+	// FormatExpiresAt; consumers must decode with ParseExpiresAt so a
+	// future format change stays a single edit.
+	MetaSessionExpiresAt = "session_expires_at"
+	// MetaSessionLeadMinutes carries the lead in whole minutes (WarningLead
+	// for the T-10 event, FinalWarningLead for the T-2 event) so the UI
+	// can show "expires in ~N minutes" without hardcoding either constant.
+	MetaSessionLeadMinutes = "lead_minutes"
+	// MetaSessionDeadlineRejected is attached to the ERROR/AUTHENTICATION
+	// SystemEvent the daemon emits when it discards a deadline from the
+	// management server (pre-epoch, too far in the future, or past the
+	// clock-skew tolerance). The value is the rejection reason string.
+	// userMessage is left empty; the UI detects the event via this key
+	// and builds a localized notification — same pattern as the session
+	// warnings above.
+	MetaSessionDeadlineRejected = "session_deadline_rejected"
+)
+
+// expiresAtLayout is the wire format used for MetaSessionExpiresAt.
+// Producer and consumers both go through FormatExpiresAt/ParseExpiresAt
+// so this layout stays a single source of truth.
+const expiresAtLayout = time.RFC3339
+
+// FormatExpiresAt encodes a deadline for MetaSessionExpiresAt. Always
+// emits UTC so a consumer in another timezone reads the same wall-clock
+// deadline.
+func FormatExpiresAt(t time.Time) string {
+	return t.UTC().Format(expiresAtLayout)
+}
+
+// ParseExpiresAt decodes the MetaSessionExpiresAt value back to a UTC
+// time. Returns an error when the field is empty or malformed; the
+// caller decides whether to fall back (zero value) or propagate.
+func ParseExpiresAt(s string) (time.Time, error) {
+	t, err := time.Parse(expiresAtLayout, s)
+	if err != nil {
+		return time.Time{}, err
+	}
+	return t.UTC(), nil
+}
+
+// FormatLeadMinutes encodes a lead duration for MetaSessionLeadMinutes
+// as the integer count of whole minutes. Sub-minute residuals are
+// truncated — the field is informational ("expires in ~N minutes") and
+// fractional minutes don't change what the UI displays.
+func FormatLeadMinutes(d time.Duration) string {
+	return strconv.Itoa(int(d / time.Minute))
+}
+
+// ParseLeadMinutes decodes a MetaSessionLeadMinutes value. Returns 0
+// and the parse error for malformed input; consumers that prefer a
+// silent fallback can simply ignore the error.
+func ParseLeadMinutes(s string) (int, error) {
+	return strconv.Atoi(s)
+}

client/internal/auth/sessionwatch/watcher.go

@@ -0,0 +1,387 @@
+// Package sessionwatch tracks the SSO session expiry deadline that the
+// management server publishes via LoginResponse / SyncResponse and fires
+// two warning events at fixed lead times before expiry: an interactive
+// T-WarningLead notification and a dismiss-gated T-FinalWarningLead
+// fallback dialog.
+//
+// The watcher is idempotent: Update may be called as often as the network
+// map snapshots arrive. Repeating the same deadline is a no-op; a new
+// deadline reschedules the timers and arms a fresh warning cycle.
+//
+// Warning firing is edge-detected. Each unique deadline value fires each
+// warning callback at most once.
+package sessionwatch
+
+import (
+	"errors"
+	"fmt"
+	"sync"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+
+	cProto "github.com/netbirdio/netbird/client/proto"
+)
+
+const (
+	// Skew tolerates a small clock difference between the management
+	// server and this peer before treating a deadline as "in the past".
+	// Slightly above typical NTP drift; tight enough that the UI doesn't
+	// paint a stale expiry as if it were valid.
+	Skew = 30 * time.Second
+
+	// maxDeadlineHorizon caps how far in the future an accepted deadline
+	// can sit. A timestamp beyond this is almost certainly a protocol
+	// glitch, and silently arming a 100-year timer would hide the bug.
+	maxDeadlineHorizon = 10 * 365 * 24 * time.Hour
+
+	// WarningLead is how far before expiry the first (interactive)
+	// warning fires. Drives the T-10 OS notification with
+	// Extend/Dismiss actions.
+	WarningLead = 10 * time.Minute
+
+	// FinalWarningLead is how far before expiry the fallback final
+	// warning fires. Drives the auto-opened SessionAboutToExpire dialog,
+	// but only when the user has not dismissed the T-WarningLead warning
+	// for the same deadline. Must be strictly less than WarningLead.
+	FinalWarningLead = 2 * time.Minute
+)
+
+var (
+	// ErrDeadlineBeforeEpoch is returned by Update when the supplied
+	// deadline pre-dates 1970-01-01.
+	ErrDeadlineBeforeEpoch = errors.New("session deadline before unix epoch")
+
+	// ErrDeadlineTooFarFuture is returned by Update when the supplied
+	// deadline is more than maxDeadlineHorizon in the future.
+	ErrDeadlineTooFarFuture = errors.New("session deadline too far in the future")
+
+	// ErrDeadlineInPast is returned by Update when the supplied deadline
+	// is more than Skew in the past.
+	ErrDeadlineInPast = errors.New("session deadline in the past")
+)
+
+// StatusRecorder is the side-effect surface the watcher drives on every
+// state transition. Production wires this to peer.Status (SetSessionExpiresAt
+// for deadline change/clear, PublishEvent for the two warnings); tests pass
+// a fake recorder so the same surface is observable without an engine.
+//
+// The watcher is the single owner of the deadline propagated to the
+// recorder: every set, clear, sanity-check rejection and Close routes the
+// value through SetSessionExpiresAt, so the SubscribeStatus snapshot the UI
+// reads can never drift from the watcher's timer state. (SetSessionExpiresAt
+// fans out its own state-change notification, so no separate notify is
+// needed.) The recorder is server-scoped and outlives this engine-scoped
+// watcher — without the Close-time clear a teardown (Down, or the Down+Up of
+// a profile switch) would leave the next session showing the previous one's
+// stale "expires in" value.
+//
+// PublishEvent's signature mirrors peer.Status.PublishEvent: the watcher
+// composes the metadata internally so the wire format (MetaSession*) is
+// owned by sessionwatch, not the caller.
+type StatusRecorder interface {
+	SetSessionExpiresAt(deadline time.Time)
+	PublishEvent(
+		severity cProto.SystemEvent_Severity,
+		category cProto.SystemEvent_Category,
+		message string,
+		userMessage string,
+		metadata map[string]string,
+	)
+}
+
+// Watcher observes the latest session deadline and fires two warnings
+// before it expires: the interactive T-WarningLead notification, and the
+// fallback T-FinalWarningLead dialog (suppressed when the user dismissed
+// the first one for the same deadline). Safe for concurrent use.
+type Watcher struct {
+	lead      time.Duration
+	finalLead time.Duration
+
+	mu           sync.Mutex
+	current      time.Time
+	timer        *time.Timer
+	finalTimer   *time.Timer
+	firedAt      time.Time // deadline value the T-WarningLead callback last fired against
+	finalFiredAt time.Time // deadline value the T-FinalWarningLead callback last fired against
+	dismissedAt  time.Time // deadline value the user dismissed via Dismiss(); gates fireFinal
+	closed       bool
+	recorder     StatusRecorder
+}
+
+// New returns a watcher with the package defaults WarningLead and
+// FinalWarningLead. Pass nil for recorder to silence side effects (handy
+// in unit tests that exercise sanity checks without observing the publish
+// path).
+func New(recorder StatusRecorder) *Watcher {
+	return NewWithLeads(WarningLead, FinalWarningLead, recorder)
+}
+
+// NewWithLeads returns a watcher with custom lead times. Useful for tests.
+// final must be strictly less than lead; otherwise both timers fire in the
+// wrong order or simultaneously and the UI flow breaks. A zero final lead
+// disables the final-warning timer entirely (see armTimerLocked) so a
+// millisecond-scale deadline doesn't flush both timers in one tick.
+func NewWithLeads(lead, final time.Duration, recorder StatusRecorder) *Watcher {
+	return &Watcher{
+		lead:      lead,
+		finalLead: final,
+		recorder:  recorder,
+	}
+}
+
+// Update sets the latest deadline. Pass the zero time to clear (e.g. when
+// a Sync push from the server omits the field because login expiration
+// was disabled).
+//
+// Same-value updates are no-ops. A different non-zero value cancels any
+// pending timer, resets the "already fired" guard, and arms a new one.
+//
+// Returns one of the sentinel Err* values when the deadline fails the
+// sanity checks (pre-epoch, far future, or in the past beyond Skew).
+// In every error case the watcher first clears its state so it stays
+// consistent with what the caller will push into its other sinks (e.g.
+// applySessionDeadline forces a zero deadline into the status recorder
+// after a non-nil error).
+func (w *Watcher) Update(deadline time.Time) error {
+	w.mu.Lock()
+	if w.closed {
+		w.mu.Unlock()
+		return nil
+	}
+
+	if deadline.IsZero() {
+		w.clearLocked()
+		return nil
+	}
+
+	now := time.Now()
+	switch {
+	case deadline.Before(time.Unix(0, 0)):
+		w.clearLocked()
+		return fmt.Errorf("%w: %v", ErrDeadlineBeforeEpoch, deadline)
+	case deadline.After(now.Add(maxDeadlineHorizon)):
+		w.clearLocked()
+		return fmt.Errorf("%w: %v", ErrDeadlineTooFarFuture, deadline)
+	case deadline.Before(now.Add(-Skew)):
+		w.clearLocked()
+		return fmt.Errorf("%w: %v (now=%v)", ErrDeadlineInPast, deadline, now)
+	}
+
+	if deadline.Equal(w.current) {
+		w.mu.Unlock()
+		return nil
+	}
+
+	w.stopTimerLocked()
+	w.current = deadline
+	// Reset every per-deadline guard so a refreshed deadline arms a fresh
+	// warning cycle: both edge triggers and the user Dismiss decision
+	// (the user agreed to the old deadline expiring; a new deadline
+	// restarts the contract).
+	w.firedAt = time.Time{}
+	w.finalFiredAt = time.Time{}
+	w.dismissedAt = time.Time{}
+
+	w.armTimerLocked(deadline)
+	recorder := w.recorder
+	w.mu.Unlock()
+	if recorder != nil {
+		recorder.SetSessionExpiresAt(deadline)
+	}
+	log.Infof("auth session deadline set to: %s (in %s)", deadline.Format(time.RFC3339), time.Until(deadline).Round(time.Second))
+	return nil
+}
+
+// Deadline returns the most recently observed deadline. Zero when no
+// deadline is currently tracked.
+func (w *Watcher) Deadline() time.Time {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	return w.current
+}
+
+// Dismiss records the user's "Dismiss" action against the current deadline
+// and suppresses the upcoming final-warning callback for that deadline.
+// Idempotent: repeated calls are no-ops. A subsequent Update with a fresh
+// deadline resets the dismissal so the final-warning cycle re-arms.
+//
+// No-op when the watcher holds no deadline or has been closed.
+func (w *Watcher) Dismiss() {
+	w.mu.Lock()
+	defer w.mu.Unlock()
+	if w.closed || w.current.IsZero() {
+		return
+	}
+	if w.dismissedAt.Equal(w.current) {
+		return
+	}
+	w.dismissedAt = w.current
+	// Cancel the armed final-warning timer eagerly. fireFinal would also
+	// gate on dismissedAt, but stopping the timer avoids a wakeup with
+	// nothing to do and makes the intent visible.
+	if w.finalTimer != nil {
+		w.finalTimer.Stop()
+		w.finalTimer = nil
+	}
+	log.Infof("auth session final-warning dismissed for deadline %s", w.current.Format(time.RFC3339))
+}
+
+// Close stops any pending timer and drops the deadline on the status
+// recorder. Update calls after Close are ignored. Clearing the recorder
+// here is what keeps a teardown (Down, or the Down+Up of a profile switch)
+// from leaving the next session showing this one's stale "expires in"
+// value — the recorder is server-scoped and outlives this engine-scoped
+// watcher, so nothing else drops the anchor on teardown.
+func (w *Watcher) Close() {
+	w.mu.Lock()
+	if w.closed {
+		w.mu.Unlock()
+		return
+	}
+	w.closed = true
+	w.stopTimerLocked()
+	hadDeadline := !w.current.IsZero()
+	w.current = time.Time{}
+	w.firedAt = time.Time{}
+	w.finalFiredAt = time.Time{}
+	w.dismissedAt = time.Time{}
+	recorder := w.recorder
+	w.mu.Unlock()
+	if recorder != nil && hadDeadline {
+		recorder.SetSessionExpiresAt(time.Time{})
+	}
+}
+
+// clearLocked drops the tracked deadline and notifies the recorder so
+// downstream consumers (SubscribeStatus stream, UI) drop their anchor.
+// The caller must hold w.mu; this helper releases it before invoking
+// the recorder.
+func (w *Watcher) clearLocked() {
+	if w.current.IsZero() {
+		w.mu.Unlock()
+		return
+	}
+	w.stopTimerLocked()
+	w.current = time.Time{}
+	w.firedAt = time.Time{}
+	w.finalFiredAt = time.Time{}
+	w.dismissedAt = time.Time{}
+	recorder := w.recorder
+	w.mu.Unlock()
+	if recorder != nil {
+		recorder.SetSessionExpiresAt(time.Time{})
+	}
+	log.Infof("auth session deadline cleared")
+}
+
+func (w *Watcher) stopTimerLocked() {
+	if w.timer != nil {
+		w.timer.Stop()
+		w.timer = nil
+	}
+	if w.finalTimer != nil {
+		w.finalTimer.Stop()
+		w.finalTimer = nil
+	}
+}
+
+func (w *Watcher) armTimerLocked(deadline time.Time) {
+	w.timer = armOneShotLocked(deadline.Add(-w.lead), func() { w.fire(deadline) })
+	// finalLead <= 0 disables the final-warning timer entirely. Used by
+	// tests that predate the final-warning fallback so a millisecond-scale
+	// deadline does not flush both timers at once.
+	if w.finalLead > 0 {
+		w.finalTimer = armOneShotLocked(deadline.Add(-w.finalLead), func() { w.fireFinal(deadline) })
+	}
+}
+
+func (w *Watcher) fire(armedFor time.Time) {
+	w.mu.Lock()
+	if w.closed || !w.current.Equal(armedFor) {
+		// Deadline moved while we were waiting (e.g. a successful extend).
+		// The reschedule path armed a fresh timer; this one is stale.
+		w.mu.Unlock()
+		return
+	}
+	if !w.firedAt.IsZero() && w.firedAt.Equal(armedFor) {
+		w.mu.Unlock()
+		return
+	}
+	w.firedAt = armedFor
+	recorder := w.recorder
+	w.mu.Unlock()
+	if recorder == nil {
+		return
+	}
+	log.Infof("auth session expiry soon warning fired")
+	publishWarning(recorder, armedFor, false)
+}
+
+// fireFinal mirrors fire for the T-FinalWarningLead timer with an extra
+// dismiss-gate: if the user dismissed the T-WarningLead notification for
+// this deadline, the final warning is suppressed entirely.
+func (w *Watcher) fireFinal(armedFor time.Time) {
+	w.mu.Lock()
+	if w.closed || !w.current.Equal(armedFor) {
+		w.mu.Unlock()
+		return
+	}
+	if !w.finalFiredAt.IsZero() && w.finalFiredAt.Equal(armedFor) {
+		w.mu.Unlock()
+		return
+	}
+	if w.dismissedAt.Equal(armedFor) {
+		w.mu.Unlock()
+		log.Infof("auth session final-warning skipped (dismissed by user)")
+		return
+	}
+	w.finalFiredAt = armedFor
+	recorder := w.recorder
+	w.mu.Unlock()
+	if recorder == nil {
+		return
+	}
+	log.Infof("auth session final-warning fired")
+	publishWarning(recorder, armedFor, true)
+}
+
+// armOneShotLocked schedules cb at fireAt. When fireAt is already in the
+// past it dispatches on the next scheduler tick so a state-change recorder
+// notification (invoked after w.mu is released) lands first. Caller must
+// hold w.mu.
+func armOneShotLocked(fireAt time.Time, cb func()) *time.Timer {
+	delay := time.Until(fireAt)
+	if delay <= 0 {
+		return time.AfterFunc(0, cb)
+	}
+	return time.AfterFunc(delay, cb)
+}
+
+// publishWarning composes the SystemEvent for a watcher-fired warning and
+// pushes it through the recorder. Severity is CRITICAL on both — bypassing
+// the user's Notifications toggle is deliberate: missing the warning
+// window forces the post-mortem SessionExpired flow (tunnel torn down,
+// lock icon, manual re-login), which is the UX we are trying to avoid.
+func publishWarning(recorder StatusRecorder, deadline time.Time, final bool) {
+	lead := WarningLead
+	message := "session expiry warning"
+	meta := map[string]string{
+		MetaSessionWarning:   "true",
+		MetaSessionExpiresAt: FormatExpiresAt(deadline),
+	}
+	if final {
+		lead = FinalWarningLead
+		message = "session expiry final warning"
+		meta[MetaSessionFinal] = "true"
+	}
+	meta[MetaSessionLeadMinutes] = FormatLeadMinutes(lead)
+
+	recorder.PublishEvent(
+		cProto.SystemEvent_CRITICAL,
+		cProto.SystemEvent_AUTHENTICATION,
+		message,
+		"",
+		meta,
+	)
+}

client/internal/auth/sessionwatch/watcher_test.go

@@ -0,0 +1,519 @@
+package sessionwatch
+
+import (
+	"errors"
+	"sync"
+	"testing"
+	"time"
+
+	cProto "github.com/netbirdio/netbird/client/proto"
+)
+
+// fakeRecorder satisfies StatusRecorder and records every call so tests
+// can observe what the watcher emits. SetSessionExpiresAt and PublishEvent
+// land in the same ordered events slice (with the Kind distinguishing
+// them) so tests that care about ordering still work. lastDeadline holds
+// the most recent value passed to SetSessionExpiresAt so tests can assert
+// the recorder ended up cleared/set as expected.
+type fakeRecorder struct {
+	mu           sync.Mutex
+	events       []event
+	lastDeadline time.Time
+}
+
+type eventKind int
+
+const (
+	stateChange eventKind = iota
+	publish
+)
+
+type event struct {
+	kind eventKind
+	// Set only for publish events.
+	severity cProto.SystemEvent_Severity
+	category cProto.SystemEvent_Category
+	message  string
+	meta     map[string]string
+}
+
+// SetSessionExpiresAt mirrors peer.Status: a same-value write is a no-op,
+// a real change records the new value and fans out a state-change (the
+// production recorder calls notifyStateChange internally). The baseline
+// is the zero time, so an initial clear before any deadline is set emits
+// nothing — matching the real recorder.
+func (r *fakeRecorder) SetSessionExpiresAt(deadline time.Time) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	if r.lastDeadline.Equal(deadline) {
+		return
+	}
+	r.lastDeadline = deadline
+	r.events = append(r.events, event{kind: stateChange})
+}
+
+func (r *fakeRecorder) deadline() time.Time {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	return r.lastDeadline
+}
+
+func (r *fakeRecorder) PublishEvent(
+	severity cProto.SystemEvent_Severity,
+	category cProto.SystemEvent_Category,
+	message string,
+	_ string,
+	metadata map[string]string,
+) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.events = append(r.events, event{
+		kind:     publish,
+		severity: severity,
+		category: category,
+		message:  message,
+		meta:     metadata,
+	})
+}
+
+func (r *fakeRecorder) snapshot() []event {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	out := make([]event, len(r.events))
+	copy(out, r.events)
+	return out
+}
+
+func (e event) isFinalWarning() bool {
+	return e.kind == publish && e.meta[MetaSessionFinal] == "true"
+}
+
+func (e event) isWarning() bool {
+	return e.kind == publish && e.meta[MetaSessionWarning] == "true" && e.meta[MetaSessionFinal] != "true"
+}
+
+func countWhere(events []event, pred func(event) bool) int {
+	n := 0
+	for _, e := range events {
+		if pred(e) {
+			n++
+		}
+	}
+	return n
+}
+
+func waitForEvents(t *testing.T, r *fakeRecorder, want int) []event {
+	t.Helper()
+	deadline := time.Now().Add(500 * time.Millisecond)
+	for time.Now().Before(deadline) {
+		if got := r.snapshot(); len(got) >= want {
+			return got
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+	got := r.snapshot()
+	t.Fatalf("timed out waiting for %d events, got %d: %+v", want, len(got), got)
+	return nil
+}
+
+// newWatcher builds a watcher with the final timer disabled (finalLead=0),
+// matching the lead-only behaviour the pre-final-warning tests assume.
+func newWatcher(lead time.Duration, r *fakeRecorder) *Watcher {
+	return NewWithLeads(lead, 0, r)
+}
+
+func TestUpdateZeroBeforeAnythingIsNoop(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	_ = w.Update(time.Time{})
+
+	if got := r.snapshot(); len(got) != 0 {
+		t.Fatalf("expected no events on initial zero, got %+v", got)
+	}
+}
+
+func TestUpdateNonZeroFiresStateChange(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	d := time.Now().Add(time.Hour)
+	_ = w.Update(d)
+
+	events := waitForEvents(t, r, 1)
+	if events[0].kind != stateChange {
+		t.Fatalf("expected stateChange, got %+v", events[0])
+	}
+	if !w.Deadline().Equal(d) {
+		t.Fatalf("deadline mismatch: %v vs %v", w.Deadline(), d)
+	}
+}
+
+func TestSameDeadlineIsNoop(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	d := time.Now().Add(time.Hour)
+	_ = w.Update(d)
+	_ = w.Update(d)
+	_ = w.Update(d)
+
+	events := waitForEvents(t, r, 1)
+	if len(events) != 1 {
+		t.Fatalf("expected exactly 1 event for repeated same deadline, got %d: %+v", len(events), events)
+	}
+}
+
+func TestWarningFiresOnceWithinLeadWindow(t *testing.T) {
+	r := &fakeRecorder{}
+	lead := 50 * time.Millisecond
+	w := newWatcher(lead, r)
+	defer w.Close()
+
+	// Deadline 80ms out — warning should fire after ~30ms.
+	d := time.Now().Add(80 * time.Millisecond)
+	_ = w.Update(d)
+
+	events := waitForEvents(t, r, 2)
+	if events[0].kind != stateChange {
+		t.Fatalf("event[0] should be stateChange, got %+v", events[0])
+	}
+	if !events[1].isWarning() {
+		t.Fatalf("event[1] should be a warning publish, got %+v", events[1])
+	}
+}
+
+func TestWarningFiresImmediatelyWhenAlreadyInsideWindow(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(time.Hour, r) // lead > delta => fire immediately
+	defer w.Close()
+
+	d := time.Now().Add(10 * time.Millisecond)
+	_ = w.Update(d)
+
+	events := waitForEvents(t, r, 2)
+	if !events[1].isWarning() {
+		t.Fatalf("expected immediate warning publish, got %+v", events[1])
+	}
+}
+
+func TestNewDeadlineCancelsPriorTimer(t *testing.T) {
+	r := &fakeRecorder{}
+	lead := 50 * time.Millisecond
+	w := newWatcher(lead, r)
+	defer w.Close()
+
+	first := time.Now().Add(80 * time.Millisecond) // would fire warning ~30ms in
+	_ = w.Update(first)
+
+	// Replace with a far-future deadline before the warning fires.
+	time.Sleep(5 * time.Millisecond)
+	second := time.Now().Add(time.Hour)
+	_ = w.Update(second)
+
+	// Wait past when first's warning would have fired.
+	time.Sleep(80 * time.Millisecond)
+
+	if n := countWhere(r.snapshot(), event.isWarning); n != 0 {
+		t.Fatalf("warning fired for cancelled deadline: %+v", r.snapshot())
+	}
+}
+
+func TestRefreshAfterFireArmsNewWarning(t *testing.T) {
+	r := &fakeRecorder{}
+	lead := 30 * time.Millisecond
+	w := newWatcher(lead, r)
+	defer w.Close()
+
+	first := time.Now().Add(50 * time.Millisecond)
+	_ = w.Update(first)
+
+	// Wait for stateChange + warning of the first cycle.
+	waitForEvents(t, r, 2)
+
+	// Simulate a successful extend: brand new deadline.
+	second := time.Now().Add(60 * time.Millisecond)
+	_ = w.Update(second)
+
+	// 4 events total: stateChange, warning (first), stateChange, warning (second).
+	events := waitForEvents(t, r, 4)
+	if events[2].kind != stateChange {
+		t.Fatalf("event[2] should be stateChange for the new deadline, got %+v", events[2])
+	}
+	if !events[3].isWarning() {
+		t.Fatalf("event[3] should be a warning publish for the new deadline, got %+v", events[3])
+	}
+}
+
+func TestUpdateZeroAfterNonZeroClearsState(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(time.Hour, r)
+	defer w.Close()
+
+	d := time.Now().Add(2 * time.Hour)
+	_ = w.Update(d)
+	waitForEvents(t, r, 1)
+
+	_ = w.Update(time.Time{})
+
+	events := waitForEvents(t, r, 2)
+	if events[1].kind != stateChange {
+		t.Fatalf("expected stateChange on clear, got %+v", events[1])
+	}
+	if !w.Deadline().IsZero() {
+		t.Fatalf("Deadline should be zero after clear")
+	}
+}
+
+func TestUpdateRejectsBeforeEpoch(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	good := time.Now().Add(time.Hour)
+	if err := w.Update(good); err != nil {
+		t.Fatalf("seed Update: %v", err)
+	}
+
+	err := w.Update(time.Unix(-100, 0))
+	if !errors.Is(err, ErrDeadlineBeforeEpoch) {
+		t.Fatalf("want ErrDeadlineBeforeEpoch, got %v", err)
+	}
+	if !w.Deadline().IsZero() {
+		t.Fatalf("rejected pre-epoch update must clear deadline; got %v", w.Deadline())
+	}
+}
+
+func TestUpdateRejectsTooFarFuture(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	good := time.Now().Add(time.Hour)
+	if err := w.Update(good); err != nil {
+		t.Fatalf("seed Update: %v", err)
+	}
+
+	err := w.Update(time.Now().Add(50 * 365 * 24 * time.Hour))
+	if !errors.Is(err, ErrDeadlineTooFarFuture) {
+		t.Fatalf("want ErrDeadlineTooFarFuture, got %v", err)
+	}
+	if !w.Deadline().IsZero() {
+		t.Fatalf("rejected far-future update must clear deadline; got %v", w.Deadline())
+	}
+}
+
+func TestUpdateInPastClearsDeadline(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	good := time.Now().Add(time.Hour)
+	if err := w.Update(good); err != nil {
+		t.Fatalf("seed Update: %v", err)
+	}
+	// Drain the stateChange from the seed.
+	waitForEvents(t, r, 1)
+
+	err := w.Update(time.Now().Add(-1 * time.Hour))
+	if !errors.Is(err, ErrDeadlineInPast) {
+		t.Fatalf("want ErrDeadlineInPast, got %v", err)
+	}
+	if !w.Deadline().IsZero() {
+		t.Fatalf("in-past update must clear the deadline, got %v", w.Deadline())
+	}
+	events := waitForEvents(t, r, 2)
+	if events[1].kind != stateChange {
+		t.Fatalf("expected stateChange on clear, got %+v", events[1])
+	}
+}
+
+func TestUpdateWithinSkewAccepted(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	// 5 seconds in the past is within the 30s Skew tolerance — accept it.
+	d := time.Now().Add(-5 * time.Second)
+	if err := w.Update(d); err != nil {
+		t.Fatalf("within-skew Update should succeed, got %v", err)
+	}
+	if !w.Deadline().Equal(d) {
+		t.Fatalf("expected deadline to be applied, got %v want %v", w.Deadline(), d)
+	}
+}
+
+func TestCloseSilencesUpdates(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	w.Close()
+
+	_ = w.Update(time.Now().Add(time.Hour))
+
+	time.Sleep(20 * time.Millisecond)
+	if got := r.snapshot(); len(got) != 0 {
+		t.Fatalf("expected no events after Close, got %+v", got)
+	}
+}
+
+// TestCloseClearsRecorderDeadline pins the profile-switch fix: a watcher
+// holding a live deadline must zero the recorder on Close so the next
+// engine's watcher (and the UI reading the shared server-scoped recorder)
+// doesn't start out showing the previous session's stale "expires in".
+func TestCloseClearsRecorderDeadline(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(time.Hour, r)
+
+	d := time.Now().Add(2 * time.Hour)
+	if err := w.Update(d); err != nil {
+		t.Fatalf("seed Update: %v", err)
+	}
+	if got := r.deadline(); !got.Equal(d) {
+		t.Fatalf("recorder deadline after Update = %v, want %v", got, d)
+	}
+
+	w.Close()
+
+	if got := r.deadline(); !got.IsZero() {
+		t.Fatalf("recorder deadline after Close = %v, want zero", got)
+	}
+}
+
+// TestCloseWithoutDeadlineLeavesRecorderUntouched guards the symmetric
+// case: closing a watcher that never held a deadline must not emit a
+// redundant clear (the recorder may legitimately hold a value written by
+// some other path; the watcher only owns what it set).
+func TestCloseWithoutDeadlineLeavesRecorderUntouched(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(time.Hour, r)
+
+	w.Close()
+
+	if got := r.snapshot(); len(got) != 0 {
+		t.Fatalf("expected no events from Close on an empty watcher, got %+v", got)
+	}
+}
+
+func TestFinalWarningFiresAfterRegularWarning(t *testing.T) {
+	r := &fakeRecorder{}
+	// Warning fires at deadline-80ms, final at deadline-30ms.
+	w := NewWithLeads(80*time.Millisecond, 30*time.Millisecond, r)
+	defer w.Close()
+
+	d := time.Now().Add(100 * time.Millisecond)
+	_ = w.Update(d)
+
+	// Expect stateChange + warning + final-warning.
+	events := waitForEvents(t, r, 3)
+
+	if countWhere(events, func(e event) bool { return e.kind == stateChange }) != 1 {
+		t.Fatalf("expected exactly 1 stateChange, got %+v", events)
+	}
+	if countWhere(events, event.isWarning) != 1 {
+		t.Fatalf("expected exactly 1 warning publish, got %+v", events)
+	}
+	if countWhere(events, event.isFinalWarning) != 1 {
+		t.Fatalf("expected exactly 1 final-warning publish, got %+v", events)
+	}
+
+	// Warning must precede final (same deadline, longer lead fires first).
+	var wIdx, fIdx int
+	for i, e := range events {
+		switch {
+		case e.isWarning():
+			wIdx = i
+		case e.isFinalWarning():
+			fIdx = i
+		}
+	}
+	if wIdx > fIdx {
+		t.Fatalf("warning must publish before final-warning, got order %+v", events)
+	}
+}
+
+func TestDismissSuppressesFinalWarning(t *testing.T) {
+	r := &fakeRecorder{}
+	w := NewWithLeads(80*time.Millisecond, 30*time.Millisecond, r)
+	defer w.Close()
+
+	d := time.Now().Add(100 * time.Millisecond)
+	_ = w.Update(d)
+
+	// Wait for the warning publish so we know we're inside the warning
+	// window, then dismiss before the final timer would fire.
+	deadline := time.Now().Add(500 * time.Millisecond)
+	for time.Now().Before(deadline) {
+		if countWhere(r.snapshot(), event.isWarning) >= 1 {
+			break
+		}
+		time.Sleep(2 * time.Millisecond)
+	}
+	if countWhere(r.snapshot(), event.isWarning) < 1 {
+		t.Fatalf("warning did not publish in time, events=%+v", r.snapshot())
+	}
+
+	w.Dismiss()
+
+	// Now wait past when the final would have fired.
+	time.Sleep(120 * time.Millisecond)
+
+	if n := countWhere(r.snapshot(), event.isFinalWarning); n != 0 {
+		t.Fatalf("final-warning published after Dismiss(), events=%+v", r.snapshot())
+	}
+}
+
+func TestDismissResetByNewDeadline(t *testing.T) {
+	r := &fakeRecorder{}
+	w := NewWithLeads(80*time.Millisecond, 30*time.Millisecond, r)
+	defer w.Close()
+
+	first := time.Now().Add(100 * time.Millisecond)
+	_ = w.Update(first)
+
+	// Dismiss against the first deadline.
+	w.Dismiss()
+
+	// Replace with a fresh deadline before the first's timers complete.
+	time.Sleep(10 * time.Millisecond)
+	second := time.Now().Add(100 * time.Millisecond)
+	_ = w.Update(second)
+
+	// The second cycle must publish a final-warning (the dismiss state
+	// did not carry over).
+	deadline := time.Now().Add(500 * time.Millisecond)
+	for time.Now().Before(deadline) {
+		if countWhere(r.snapshot(), event.isFinalWarning) >= 1 {
+			break
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+	if countWhere(r.snapshot(), event.isFinalWarning) < 1 {
+		t.Fatalf("final-warning did not publish on fresh deadline after Dismiss reset, events=%+v", r.snapshot())
+	}
+}
+
+func TestDismissBeforeUpdateIsNoop(t *testing.T) {
+	r := &fakeRecorder{}
+	w := NewWithLeads(80*time.Millisecond, 30*time.Millisecond, r)
+	defer w.Close()
+
+	// No deadline tracked yet; Dismiss must be a no-op (no panic, no state).
+	w.Dismiss()
+
+	d := time.Now().Add(100 * time.Millisecond)
+	_ = w.Update(d)
+
+	// Final warning should still publish — Dismiss only acts on the current
+	// deadline, and there was none at the time of the call.
+	deadline := time.Now().Add(500 * time.Millisecond)
+	for time.Now().Before(deadline) {
+		if countWhere(r.snapshot(), event.isFinalWarning) >= 1 {
+			return
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+	t.Fatalf("final-warning did not publish after no-op pre-Update Dismiss, events=%+v", r.snapshot())
+}

client/internal/conn_mgr.go

@@ -109,7 +109,7 @@ func (e *ConnMgr) UpdatedRemoteFeatureFlag(ctx context.Context, enabled bool) er
 			return nil
 		}
 
-		log.Warnf("lazy connection manager is enabled by management feature flag")
+		log.Infof("lazy connection manager is enabled by the management feature flag")
 		e.initLazyManager(ctx)
 		e.statusRecorder.UpdateLazyConnection(true)
 		return e.addPeersToLazyConnManager()

client/internal/connect.go

@@ -277,6 +277,15 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 		log.Debugf("connecting to the Management service %s", c.config.ManagementURL.Host)
 		mgmClient, err := mgm.NewClient(engineCtx, c.config.ManagementURL.Host, myPrivateKey, mgmTlsEnabled)
 		if err != nil {
+			// On daemon shutdown / Down() the parent context is cancelled
+			// and the dial fails with "context canceled". Wrapping that
+			// into state would leave the snapshot stuck at Connecting+err
+			// until the backoff loop wakes up — instead let the operation
+			// return cleanly so the deferred state.Set(StatusIdle) takes
+			// effect on the next iteration.
+			if c.ctx.Err() != nil {
+				return nil
+			}
 			return wrapErr(gstatus.Errorf(codes.FailedPrecondition, "failed connecting to Management Service : %s", err))
 		}
 		mgmNotifier := statusRecorderToMgmConnStateNotifier(c.statusRecorder)
@@ -415,6 +424,10 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			return wrapErr(err)
 		}
 
+		// Seed the session-expiry deadline from the LoginResponse. Subsequent
+		// changes flow in through SyncResponse and are applied in handleSync.
+		engine.ApplySessionDeadline(loginResp.GetSessionExpiresAt())
+
 		log.Infof("Netbird engine started, the IP is: %s", peerConfig.GetAddress())
 		state.Set(StatusConnected)
 
@@ -451,6 +464,10 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 	}
 
 	c.statusRecorder.ClientStart()
+	// Wrap the backoff with c.ctx so Down()/actCancel propagates into the
+	// inter-attempt sleep — otherwise a 15s MaxInterval can keep the retry
+	// loop alive long after the caller asked to give up, leaving the
+	// status stream stuck at Connecting.
 	err = backoff.Retry(operation, backoff.WithContext(backOff, c.ctx))
 	if err != nil {
 		log.Debugf("exiting client retry loop due to unrecoverable error: %s", err)

client/internal/debug/debug.go

@@ -229,9 +229,16 @@ scutil_dns.txt (macOS only):
 
 const (
 	clientLogFile = "client.log"
+	uiLogFile     = "gui-client.log"
 	errorLogFile  = "netbird.err"
 	stdoutLogFile = "netbird.out"
 
+	// Rotated-log glob prefixes (base log name without extension) passed to
+	// addRotatedLogFiles. The daemon's own log and the GUI log live in the same
+	// dir, so the prefixes must be disjoint to keep their rotated siblings apart.
+	clientLogPrefix = "client"
+	uiLogPrefix     = "gui-client"
+
 	darwinErrorLogPath  = "/var/log/netbird.out.log"
 	darwinStdoutLogPath = "/var/log/netbird.err.log"
 )
@@ -249,6 +256,7 @@ type BundleGenerator struct {
 	statusRecorder *peer.Status
 	syncResponse   *mgmProto.SyncResponse
 	logPath        string
+	uiLogPath      string
 	tempDir        string
 	statePath      string
 	cpuProfile     []byte
@@ -276,6 +284,7 @@ type GeneratorDependencies struct {
 	StatusRecorder *peer.Status
 	SyncResponse   *mgmProto.SyncResponse
 	LogPath        string
+	UILogPath      string // Absolute path to the desktop UI's gui-client.log, reported via RegisterUILog. Empty if no UI registered one.
 	TempDir        string // Directory for temporary bundle zip files. If empty, os.TempDir() is used.
 	StatePath      string // Path to the state file. If empty, the ServiceManager default path is used.
 	CPUProfile     []byte
@@ -300,6 +309,7 @@ func NewBundleGenerator(deps GeneratorDependencies, cfg BundleConfig) *BundleGen
 		statusRecorder: deps.StatusRecorder,
 		syncResponse:   deps.SyncResponse,
 		logPath:        deps.LogPath,
+		uiLogPath:      deps.UILogPath,
 		tempDir:        deps.TempDir,
 		statePath:      deps.StatePath,
 		cpuProfile:     deps.CPUProfile,
@@ -411,6 +421,10 @@ func (g *BundleGenerator) createArchive() error {
 		log.Errorf("failed to add logs to debug bundle: %v", err)
 	}
 
+	if err := g.addUILog(); err != nil {
+		log.Errorf("failed to add UI log to debug bundle: %v", err)
+	}
+
 	if err := g.addUpdateLogs(); err != nil {
 		log.Errorf("failed to add updater logs: %v", err)
 	}
@@ -986,7 +1000,7 @@ func (g *BundleGenerator) addLogfile() error {
 		return fmt.Errorf("add client log file to zip: %w", err)
 	}
 
-	g.addRotatedLogFiles(logDir)
+	g.addRotatedLogFiles(logDir, clientLogPrefix)
 
 	stdErrLogPath := filepath.Join(logDir, errorLogFile)
 	stdoutLogPath := filepath.Join(logDir, stdoutLogFile)
@@ -1006,6 +1020,25 @@ func (g *BundleGenerator) addLogfile() error {
 	return nil
 }
 
+// addUILog adds the desktop UI's gui-client.log (and its rotated siblings) to
+// the bundle. The path is reported by the UI via RegisterUILog; empty when no
+// UI registered one (e.g. headless / server). Missing file is non-fatal — the
+// UI only writes it while the daemon is in debug, so it's often absent.
+func (g *BundleGenerator) addUILog() error {
+	if g.uiLogPath == "" {
+		log.Debugf("no UI log path registered, skipping in debug bundle")
+		return nil
+	}
+
+	if err := g.addSingleLogfile(g.uiLogPath, uiLogFile); err != nil {
+		return fmt.Errorf("add UI log file to zip: %w", err)
+	}
+
+	g.addRotatedLogFiles(filepath.Dir(g.uiLogPath), uiLogPrefix)
+
+	return nil
+}
+
 // addSingleLogfile adds a single log file to the archive
 func (g *BundleGenerator) addSingleLogfile(logPath, targetName string) error {
 	logFile, err := os.Open(logPath)
@@ -1078,14 +1111,16 @@ func (g *BundleGenerator) addSingleLogFileGz(logPath, targetName string) error {
 	return nil
 }
 
-// addRotatedLogFiles adds rotated log files to the bundle based on logFileCount
-func (g *BundleGenerator) addRotatedLogFiles(logDir string) {
+// addRotatedLogFiles adds rotated log files to the bundle based on logFileCount.
+// prefix is the base log name without extension (e.g. "client", "gui-client");
+// the glob matches both files rotated by us and by logrotate on linux.
+func (g *BundleGenerator) addRotatedLogFiles(logDir, prefix string) {
 	if g.logFileCount == 0 {
 		return
 	}
 
-	// This regex will match both logs rotated by us and logrotate on linux
-	pattern := filepath.Join(logDir, "client*.log.*")
+	// This pattern matches both logs rotated by us and logrotate on linux
+	pattern := filepath.Join(logDir, prefix+"*.log.*")
 	files, err := filepath.Glob(pattern)
 	if err != nil {
 		log.Warnf("failed to glob rotated logs: %v", err)

client/internal/debug/debug_logfiles_test.go

@@ -40,6 +40,25 @@ func TestAddRotatedLogFiles_PicksUpAllVariants(t *testing.T) {
 	require.NotContains(t, names, "other.log", "unrelated files should not be in bundle")
 }
 
+// TestAddRotatedLogFiles_GUIPrefix asserts the prefix parameter scopes the glob
+// to the GUI log: gui-client.log.* rotated siblings are picked up and the
+// daemon's own client.log.* are not (and vice versa, covered above). This is
+// the load-bearing check for the gui-client.log bundle collection — the old
+// "client*.log.*" glob would have missed gui-client rotations.
+func TestAddRotatedLogFiles_GUIPrefix(t *testing.T) {
+	dir := t.TempDir()
+
+	writeFile(t, filepath.Join(dir, "gui-client.log.1"), "gui rotated\n")
+	writeGzFile(t, filepath.Join(dir, "gui-client.log.2.gz"), "gui rotated gz\n")
+	writeFile(t, filepath.Join(dir, "client.log.1"), "daemon rotated\n")
+
+	names := runAddRotatedLogFilesPrefix(t, dir, "gui-client", 10)
+
+	require.Contains(t, names, "gui-client.log.1", "gui-client rotated file should be in bundle")
+	require.Contains(t, names, "gui-client.log.2.gz", "gui-client gz rotated file should be in bundle")
+	require.NotContains(t, names, "client.log.1", "daemon rotated file must not match the gui-client prefix")
+}
+
 // TestAddRotatedLogFiles_RespectsLogFileCount asserts that only the newest
 // logFileCount rotated files are bundled, ordered by mtime.
 func TestAddRotatedLogFiles_RespectsLogFileCount(t *testing.T) {
@@ -67,6 +86,10 @@ func TestAddRotatedLogFiles_RespectsLogFileCount(t *testing.T) {
 // runAddRotatedLogFiles calls addRotatedLogFiles against a fresh in-memory
 // zip writer and returns the set of entry names that ended up in the archive.
 func runAddRotatedLogFiles(t *testing.T, dir string, logFileCount uint32) map[string]struct{} {
+	return runAddRotatedLogFilesPrefix(t, dir, "client", logFileCount)
+}
+
+func runAddRotatedLogFilesPrefix(t *testing.T, dir, prefix string, logFileCount uint32) map[string]struct{} {
 	t.Helper()
 
 	var buf bytes.Buffer
@@ -74,7 +97,7 @@ func runAddRotatedLogFiles(t *testing.T, dir string, logFileCount uint32) map[st
 		archive:      zip.NewWriter(&buf),
 		logFileCount: logFileCount,
 	}
-	g.addRotatedLogFiles(dir)
+	g.addRotatedLogFiles(dir, prefix)
 	require.NoError(t, g.archive.Close())
 
 	zr, err := zip.NewReader(bytes.NewReader(buf.Bytes()), int64(buf.Len()))

client/internal/engine.go

@@ -48,6 +48,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/internal/peer/guard"
 	icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
 	"github.com/netbirdio/netbird/client/internal/peerstore"
 	"github.com/netbirdio/netbird/client/internal/portforward"
 	"github.com/netbirdio/netbird/client/internal/profilemanager"
@@ -182,7 +183,7 @@ type EngineServices struct {
 type Engine struct {
 	// signal is a Signal Service client
 	signal   signal.Client
-	signaler *peer.Signaler
+	signaler *signaling.Signaler
 	// mgmClient is a Management Service client
 	mgmClient mgm.Client
 	// peerConns is a map that holds all the peers that are known to this peer
@@ -220,12 +221,6 @@ type Engine struct {
 	// networkSerial is the latest CurrentSerial (state ID) of the network sent by the Management service
 	networkSerial uint64
 
-	// forwardingRules holds the ingress forward rules applied for the current target.
-	// Wholesale sections (incl. forward rules) run only on the first pass of a target;
-	// it is stashed here so the final, peer-converged pass can build the lazy-connection
-	// exclude list without recomputing them on every bounded peer pass.
-	forwardingRules []firewallManager.ForwardRule
-
 	networkMonitor *networkmonitor.NetworkMonitor
 
 	sshServer sshServer
@@ -280,6 +275,20 @@ type Engine struct {
 	jobExecutorWG sync.WaitGroup
 
 	exposeManager *expose.Manager
+
+	sessionWatcher sessionDeadlineWatcher
+}
+
+// sessionDeadlineWatcher is the engine-facing surface of the SSO session
+// expiry watcher. The concrete implementation (sessionwatch.Watcher) is wired
+// in via newSessionWatcher, which is build-tagged so the js/wasm build links a
+// no-op stub instead of pulling the full sessionwatch package (and its timer
+// machinery) into the binary — the wasm client never runs the engine's
+// session-warning flow.
+type sessionDeadlineWatcher interface {
+	Update(deadline time.Time) error
+	Dismiss()
+	Close()
 }
 
 // Peer is an instance of the Connection Peer
@@ -310,7 +319,7 @@ func NewEngine(
 		ctx:                ctx,
 		cancel:             cancel,
 		signal:             services.SignalClient,
-		signaler:           peer.NewSignaler(services.SignalClient, config.WgPrivateKey),
+		signaler:           signaling.NewSignaler(services.SignalClient, config.WgPrivateKey),
 		mgmClient:          services.MgmClient,
 		relayManager:       services.RelayManager,
 		peerStore:          peerstore.NewConnStore(),
@@ -331,6 +340,17 @@ func NewEngine(
 		updateManager:      services.UpdateManager,
 		syncStoreDir:       config.StateDir,
 	}
+	// sessionWatcher keeps the SubscribeStatus consumers in sync with the
+	// session expiry deadline. Deadline-change ticks come for free via
+	// Status.SetSessionExpiresAt; the watcher exists to push a wake-up at
+	// T-WarningLead and T-FinalWarningLead so the UI repaints the remaining
+	// time / warning state even when nothing else changed, and to publish
+	// two SystemEvents (the warning composition lives in sessionwatch so
+	// the wire format stays owned by one package):
+	//   - T-WarningLead   → interactive "Extend now / Dismiss" notification
+	//   - T-FinalWarningLead → auto-opened SessionAboutToExpire dialog,
+	//     suppressed when the user dismissed the earlier warning
+	engine.sessionWatcher = newSessionWatcher(engine.statusRecorder)
 
 	log.Infof("I am: %s", config.WgPrivateKey.PublicKey().String())
 	return engine
@@ -397,6 +417,10 @@ func (e *Engine) stopLocked() {
 		e.srWatcher.Close()
 	}
 
+	if e.sessionWatcher != nil {
+		e.sessionWatcher.Close()
+	}
+
 	if e.updateManager != nil {
 		e.updateManager.SetDownloadOnly()
 	}
@@ -528,7 +552,7 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 		} else {
 			log.Infof("running rosenpass in strict mode")
 		}
-		e.rpManager, err = rosenpass.NewManager(e.config.PreSharedKey, e.config.WgIfaceName)
+		e.rpManager, err = rosenpass.NewManager(e.config.PreSharedKey, e.config.WgIfaceName, publicKey)
 		if err != nil {
 			return fmt.Errorf("create rosenpass manager: %w", err)
 		}
@@ -780,15 +804,7 @@ func (e *Engine) blockLanAccess() {
 
 // modifyPeers updates peers that have been modified (e.g. IP address has been changed).
 // It closes the existing connection, removes it from the peerConns map, and creates a new one.
-// maxPeersPerSyncPass is the default per-pass cap on how many peers each of
-// removePeers/modifyPeers/addNewPeers applies, so syncMsgMux is held only for a
-// batch at a time and other subsystems can interleave between passes. It is
-// passed in (not read globally) so tests can exercise the multi-pass path.
-const maxPeersPerSyncPass = 300
-
-// modifyPeers re-applies up to maxBatch changed peers per call. It returns true
-// when more changed peers remained than the cap, so the caller re-runs.
-func (e *Engine) modifyPeers(peersUpdate []*mgmProto.RemotePeerConfig, maxBatch int) (bool, error) {
+func (e *Engine) modifyPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 
 	// first, check if peers have been modified
 	var modified []*mgmProto.RemotePeerConfig
@@ -818,32 +834,26 @@ func (e *Engine) modifyPeers(peersUpdate []*mgmProto.RemotePeerConfig, maxBatch
 		}
 	}
 
-	more := false
-	if len(modified) > maxBatch {
-		modified = modified[:maxBatch]
-		more = true
-	}
-
 	// second, close all modified connections and remove them from the state map
 	for _, p := range modified {
-		if err := e.removePeer(p.GetWgPubKey()); err != nil {
-			return false, err
+		err := e.removePeer(p.GetWgPubKey())
+		if err != nil {
+			return err
 		}
 	}
 	// third, add the peer connections again
 	for _, p := range modified {
-		if err := e.addNewPeer(p); err != nil {
-			return false, err
+		err := e.addNewPeer(p)
+		if err != nil {
+			return err
 		}
 	}
-	return more, nil
+	return nil
 }
 
 // removePeers finds and removes peers that do not exist anymore in the network map received from the Management Service.
 // It also removes peers that have been modified (e.g. change of IP address). They will be added again in addPeers method.
-// removePeers removes up to maxBatch peers per call. It returns true when more
-// peers remained to remove than the cap, so the caller re-runs.
-func (e *Engine) removePeers(peersUpdate []*mgmProto.RemotePeerConfig, maxBatch int) (bool, error) {
+func (e *Engine) removePeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	newPeers := make([]string, 0, len(peersUpdate))
 	for _, p := range peersUpdate {
 		newPeers = append(newPeers, p.GetWgPubKey())
@@ -851,19 +861,14 @@ func (e *Engine) removePeers(peersUpdate []*mgmProto.RemotePeerConfig, maxBatch
 
 	toRemove := util.SliceDiff(e.peerStore.PeersPubKey(), newPeers)
 
-	more := false
-	if len(toRemove) > maxBatch {
-		toRemove = toRemove[:maxBatch]
-		more = true
-	}
-
 	for _, p := range toRemove {
-		if err := e.removePeer(p); err != nil {
-			return false, err
+		err := e.removePeer(p)
+		if err != nil {
+			return err
 		}
 		log.Infof("removed peer %s", p)
 	}
-	return more, nil
+	return nil
 }
 
 func (e *Engine) removeAllPeers() error {
@@ -942,19 +947,23 @@ func (e *Engine) phase(name string) func() {
 	}
 }
 
-// applySyncPass applies one bounded pass of the sync update under syncMsgMux and
-// returns true if more peers remained than the per-pass cap. It is driven by the
-// mapStateManager, which re-invokes it (releasing the lock between passes) until
-// the update is fully applied.
-func (e *Engine) applySyncPass(update *mgmProto.SyncResponse, firstPass bool) (bool, error) {
+func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
+	started := time.Now()
+	defer func() {
+		duration := time.Since(started)
+		log.Infof("sync finished in %s", duration)
+		e.clientMetrics.RecordSyncDuration(e.ctx, duration)
+	}()
 	e.syncMsgMux.Lock()
 	defer e.syncMsgMux.Unlock()
 
 	// Check context INSIDE lock to ensure atomicity with shutdown
 	if e.ctx.Err() != nil {
-		return false, e.ctx.Err()
+		return e.ctx.Err()
 	}
 
+	e.ApplySessionDeadline(update.GetSessionExpiresAt())
+
 	if update.NetworkMap != nil && update.NetworkMap.PeerConfig != nil {
 		e.handleAutoUpdateVersion(update.NetworkMap.PeerConfig.AutoUpdate)
 	}
@@ -963,7 +972,7 @@ func (e *Engine) applySyncPass(update *mgmProto.SyncResponse, firstPass bool) (b
 	err := e.updateNetbirdConfig(update.GetNetbirdConfig())
 	done()
 	if err != nil {
-		return false, err
+		return err
 	}
 
 	// Posture checks are bound to the network map presence:
@@ -973,25 +982,28 @@ func (e *Engine) applySyncPass(update *mgmProto.SyncResponse, firstPass bool) (b
 	//                                        leave the previously applied checks untouched
 	nm := update.GetNetworkMap()
 	if nm == nil {
-		return false, nil
+		return nil
 	}
 
 	done = e.phase("checks")
 	err = e.updateChecksIfNew(update.Checks)
 	done()
 	if err != nil {
-		return false, err
+		return err
 	}
 
+	done = e.phase("persist")
+	e.persistSyncResponse(update)
+	done()
+
 	// only apply new changes and ignore old ones
-	more, err := e.updateNetworkMap(nm, maxPeersPerSyncPass, firstPass)
-	if err != nil {
-		return false, err
+	if err := e.updateNetworkMap(nm); err != nil {
+		return err
 	}
 
 	e.statusRecorder.PublishEvent(cProto.SystemEvent_INFO, cProto.SystemEvent_SYSTEM, "Network map updated", "", nil)
 
-	return more, nil
+	return nil
 }
 
 // updateNetbirdConfig applies the management-provided NetBird configuration:
@@ -1039,13 +1051,6 @@ func (e *Engine) updateNetbirdConfig(wCfg *mgmProto.NetbirdConfig) error {
 // (not syncMsgMux) is held for the whole Set so the store cannot be cleared (disabled /
 // engine close) mid-call and have this write resurrect a file that was just removed.
 func (e *Engine) persistSyncResponse(update *mgmProto.SyncResponse) {
-	// Only persist updates that carry a network map. Config-only updates (e.g. relay
-	// token rotation, STUN/TURN) have a nil NetworkMap; persisting them would overwrite
-	// the last full map on disk and break restore-on-restart.
-	if update.GetNetworkMap() == nil {
-		return
-	}
-
 	e.syncRespMux.RLock()
 	defer e.syncRespMux.RUnlock()
 
@@ -1294,7 +1299,7 @@ func (e *Engine) handleBundle(params *mgmProto.BundleParameters) (*mgmProto.JobR
 		ClientMetrics:  e.clientMetrics,
 		DaemonVersion:  version.NetbirdVersion(),
 		RefreshStatus: func() {
-			e.RunHealthProbes(true)
+			e.RunHealthProbes(e.ctx, true)
 		},
 	}
 
@@ -1333,24 +1338,7 @@ func (e *Engine) receiveManagementEvents() {
 		}
 		e.applyInfoFlags(info)
 
-		// The map-state manager converges the latest update in the background in
-		// bounded passes; the stream callback only hands it the newest target.
-		persist := func(u *mgmProto.SyncResponse) {
-			done := e.phase("persist")
-			e.persistSyncResponse(u)
-			done()
-		}
-		manager := newMapStateManager(e.applySyncPass, persist, func(d time.Duration) {
-			log.Infof("sync finished in %s", d)
-			e.clientMetrics.RecordSyncDuration(e.ctx, d)
-		})
-		e.shutdownWg.Add(1)
-		go func() {
-			defer e.shutdownWg.Done()
-			manager.run(e.ctx)
-		}()
-
-		err := e.mgmClient.Sync(e.ctx, info, manager.SetTarget)
+		err := e.mgmClient.Sync(e.ctx, info, e.handleSync)
 		if err != nil {
 			// happens if management is unavailable for a long time.
 			// We want to cancel the operation of the whole client
@@ -1401,107 +1389,21 @@ func (e *Engine) updateTURNs(turns []*mgmProto.ProtectedHostConfig) error {
 	return nil
 }
 
-// updateNetworkMap applies the wholesale parts (config, routes, ACL, DNS) in full
-// and up to maxBatch peers per phase. It returns true when more peers remained
-// than the cap, so the caller re-runs until convergence.
-func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap, maxBatch int, firstPass bool) (bool, error) {
+func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 	// intentionally leave it before checking serial because for now it can happen that peer IP changed but serial didn't
 	if networkMap.GetPeerConfig() != nil {
 		err := e.updateConfig(networkMap.GetPeerConfig())
 		if err != nil {
-			return false, err
+			return err
 		}
 	}
 
 	serial := networkMap.GetSerial()
 	if e.networkSerial > serial {
 		log.Debugf("received outdated NetworkMap with serial %d, ignoring", serial)
-		return false, nil
+		return nil
 	}
 
-	// Wholesale sections (firewall/ACL, DNS, routes, forward rules) are applied
-	// up-front and only once per target: they are cheap, local, idempotent and must
-	// be in place before peers come up (fail-closed). On the bounded re-runs that only
-	// drain the remaining peer batches they are skipped — the applied forward rules are
-	// reused from e.forwardingRules for the lazy-exclude finalize.
-	if firstPass {
-		e.applyWholesale(networkMap, serial)
-	}
-
-	log.Debugf("got peers update from Management Service, total peers to connect to = %d", len(networkMap.GetRemotePeers()))
-
-	doneOffline := e.phase("offline_peers")
-	e.updateOfflinePeers(networkMap.GetOfflinePeers())
-	doneOffline()
-
-	// Filter out own peer from the remote peers list
-	localPubKey := e.config.WgPrivateKey.PublicKey().String()
-	remotePeers := make([]*mgmProto.RemotePeerConfig, 0, len(networkMap.GetRemotePeers()))
-	for _, p := range networkMap.GetRemotePeers() {
-		if p.GetWgPubKey() != localPubKey {
-			remotePeers = append(remotePeers, p)
-		}
-	}
-
-	// No special case for cleanup: when management signals RemotePeersIsEmpty (e.g. our
-	// peer was deleted), remotePeers is already empty, so the bounded diff below removes
-	// every peer in batches — same path as a normal update, no unbounded removeAllPeers
-	// held under syncMsgMux in one shot.
-	doneRemoved := e.phase("removed_peers")
-	removeMore, err := e.removePeers(remotePeers, maxBatch)
-	doneRemoved()
-	if err != nil {
-		return false, err
-	}
-
-	doneModified := e.phase("modified_peers")
-	modifyMore, err := e.modifyPeers(remotePeers, maxBatch)
-	doneModified()
-	if err != nil {
-		return false, err
-	}
-
-	doneAdded := e.phase("added_peers")
-	addMore, err := e.addNewPeers(remotePeers, maxBatch)
-	doneAdded()
-	if err != nil {
-		return false, err
-	}
-
-	// needMore signals the caller to re-run when a peer phase hit its per-pass cap.
-	needMore := removeMore || modifyMore || addMore
-
-	e.statusRecorder.FinishPeerListModifications()
-
-	e.updatePeerSSHHostKeys(remotePeers)
-
-	if err := e.updateSSHClientConfig(remotePeers); err != nil {
-		log.Warnf("failed to update SSH client config: %v", err)
-	}
-
-	e.updateSSHServerAuth(networkMap.GetSshAuth())
-
-	// Set the exclude list only once peers have fully converged (this pass added
-	// the last batch). It needs all target peers present in the store, and
-	// ExcludePeer has replace-semantics — a partial set mid-convergence would be wrong.
-	if !needMore {
-		doneLazy := e.phase("lazy_exclude")
-		excludedLazyPeers := e.toExcludedLazyPeers(e.forwardingRules, remotePeers)
-		e.connMgr.SetExcludeList(e.ctx, excludedLazyPeers)
-		doneLazy()
-	}
-
-	e.networkSerial = serial
-
-	return needMore, nil
-}
-
-// applyWholesale applies the cheap, local, idempotent map sections — lazy feature
-// flag, firewall/legacy management, DNS, routes, ACL filtering, DNS forwarder and
-// ingress forward rules — that must be in place before peers come up. It runs once
-// per target (first pass only); the resulting forward rules are stashed in
-// e.forwardingRules for the lazy-exclude finalize on the peer-converged pass.
-func (e *Engine) applyWholesale(networkMap *mgmProto.NetworkMap, serial uint64) {
 	if err := e.connMgr.UpdatedRemoteFeatureFlag(e.ctx, networkMap.GetPeerConfig().GetLazyConnectionEnabled()); err != nil {
 		log.Errorf("failed to update lazy connection feature flag: %v", err)
 	}
@@ -1574,7 +1476,84 @@ func (e *Engine) applyWholesale(networkMap *mgmProto.NetworkMap, serial uint64)
 		log.Errorf("failed to update forward rules, err: %v", err)
 	}
 	done()
-	e.forwardingRules = forwardingRules
+
+	log.Debugf("got peers update from Management Service, total peers to connect to = %d", len(networkMap.GetRemotePeers()))
+
+	done = e.phase("offline_peers")
+	e.updateOfflinePeers(networkMap.GetOfflinePeers())
+	done()
+
+	remotePeers, err := e.reconcilePeers(networkMap)
+	if err != nil {
+		return err
+	}
+
+	// must set the exclude list after the peers are added. Without it the manager can not figure out the peers parameters from the store
+	done = e.phase("lazy_exclude")
+	excludedLazyPeers := e.toExcludedLazyPeers(forwardingRules, remotePeers)
+	e.connMgr.SetExcludeList(e.ctx, excludedLazyPeers)
+	done()
+
+	e.networkSerial = serial
+
+	return nil
+}
+
+// reconcilePeers applies the remote peer list from the network map (removing,
+// modifying and adding peers, then updating SSH config) and returns the remote
+// peers with our own peer filtered out, for use by later sync steps.
+func (e *Engine) reconcilePeers(networkMap *mgmProto.NetworkMap) ([]*mgmProto.RemotePeerConfig, error) {
+	// Filter out own peer from the remote peers list
+	localPubKey := e.config.WgPrivateKey.PublicKey().String()
+	remotePeers := make([]*mgmProto.RemotePeerConfig, 0, len(networkMap.GetRemotePeers()))
+	for _, p := range networkMap.GetRemotePeers() {
+		if p.GetWgPubKey() != localPubKey {
+			remotePeers = append(remotePeers, p)
+		}
+	}
+
+	// cleanup request, most likely our peer has been deleted
+	if networkMap.GetRemotePeersIsEmpty() {
+		err := e.removeAllPeers()
+		e.statusRecorder.FinishPeerListModifications()
+		if err != nil {
+			return nil, err
+		}
+		return remotePeers, nil
+	}
+
+	done := e.phase("removed_peers")
+	err := e.removePeers(remotePeers)
+	done()
+	if err != nil {
+		return nil, err
+	}
+
+	done = e.phase("modified_peers")
+	err = e.modifyPeers(remotePeers)
+	done()
+	if err != nil {
+		return nil, err
+	}
+
+	done = e.phase("added_peers")
+	err = e.addNewPeers(remotePeers)
+	done()
+	if err != nil {
+		return nil, err
+	}
+
+	e.statusRecorder.FinishPeerListModifications()
+
+	e.updatePeerSSHHostKeys(remotePeers)
+
+	if err := e.updateSSHClientConfig(remotePeers); err != nil {
+		log.Warnf("failed to update SSH client config: %v", err)
+	}
+
+	e.updateSSHServerAuth(networkMap.GetSshAuth())
+
+	return remotePeers, nil
 }
 
 func toDNSFeatureFlag(networkMap *mgmProto.NetworkMap) bool {
@@ -1754,23 +1733,14 @@ func addrToString(addr netip.Addr) string {
 }
 
 // addNewPeers adds peers that were not know before but arrived from the Management service with the update
-// addNewPeers adds up to maxBatch not-yet-present peers per call. It returns true
-// when more new peers remained than the cap, so the caller re-runs.
-func (e *Engine) addNewPeers(peersUpdate []*mgmProto.RemotePeerConfig, maxBatch int) (bool, error) {
-	added := 0
+func (e *Engine) addNewPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	for _, p := range peersUpdate {
-		if _, ok := e.peerStore.PeerConn(p.GetWgPubKey()); ok {
-			continue // already present (cheap skip), does not count toward the cap
+		err := e.addNewPeer(p)
+		if err != nil {
+			return err
 		}
-		if added >= maxBatch {
-			return true, nil // at least one more new peer remains
-		}
-		if err := e.addNewPeer(p); err != nil {
-			return false, err
-		}
-		added++
 	}
-	return false, nil
+	return nil
 }
 
 // addNewPeer add peer if connection doesn't exist
@@ -2255,7 +2225,20 @@ func (e *Engine) getRosenpassAddr() string {
 
 // RunHealthProbes executes health checks for Signal, Management, Relay, and WireGuard services
 // and updates the status recorder with the latest states.
-func (e *Engine) RunHealthProbes(waitForResult bool) bool {
+//
+// ctx scopes the (potentially slow) STUN/TURN probing: a caller that gives up —
+// e.g. a Status RPC whose client disconnected — cancels its ctx and the probe
+// returns instead of running to its per-component timeout. The engine's own
+// lifetime ctx still applies independently, so an engine shutdown aborts the
+// probe even if the caller's ctx is context.Background().
+func (e *Engine) RunHealthProbes(ctx context.Context, waitForResult bool) bool {
+	// Tie the caller's ctx to the engine lifetime: either cancelling aborts
+	// the probe below.
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+	stop := context.AfterFunc(e.ctx, cancel)
+	defer stop()
+
 	e.syncMsgMux.Lock()
 
 	signalHealthy := e.signal.IsHealthy()
@@ -2278,9 +2261,9 @@ func (e *Engine) RunHealthProbes(waitForResult bool) bool {
 	if runtime.GOOS != "js" {
 		var results []relay.ProbeResult
 		if waitForResult {
-			results = e.probeStunTurn.ProbeAllWaitResult(e.ctx, stuns, turns)
+			results = e.probeStunTurn.ProbeAllWaitResult(ctx, stuns, turns)
 		} else {
-			results = e.probeStunTurn.ProbeAll(e.ctx, stuns, turns)
+			results = e.probeStunTurn.ProbeAll(ctx, stuns, turns)
 		}
 		e.statusRecorder.UpdateRelayStates(results)
 
@@ -2765,7 +2748,7 @@ func createFile(path string) error {
 	return file.Close()
 }
 
-func convertToOfferAnswer(msg *sProto.Message) (*peer.OfferAnswer, error) {
+func convertToOfferAnswer(msg *sProto.Message) (*signaling.OfferAnswer, error) {
 	remoteCred, err := signal.UnMarshalCredential(msg)
 	if err != nil {
 		return nil, err
@@ -2781,9 +2764,9 @@ func convertToOfferAnswer(msg *sProto.Message) (*peer.OfferAnswer, error) {
 	}
 
 	// Handle optional SessionID
-	var sessionID *peer.ICESessionID
+	var sessionID *icemaker.SessionID
 	if sessionBytes := msg.GetBody().GetSessionId(); sessionBytes != nil {
-		if id, err := peer.ICESessionIDFromBytes(sessionBytes); err != nil {
+		if id, err := icemaker.SessionIDFromBytes(sessionBytes); err != nil {
 			log.Warnf("Invalid session ID in message: %v", err)
 			sessionID = nil // Set to nil if conversion fails
 		} else {
@@ -2793,8 +2776,8 @@ func convertToOfferAnswer(msg *sProto.Message) (*peer.OfferAnswer, error) {
 
 	relayIP := decodeRelayIP(msg.GetBody().GetRelayServerIP())
 
-	offerAnswer := peer.OfferAnswer{
-		IceCredentials: peer.IceCredentials{
+	offerAnswer := signaling.OfferAnswer{
+		IceCredentials: signaling.IceCredentials{
 			UFrag: remoteCred.UFrag,
 			Pwd:   remoteCred.Pwd,
 		},

client/internal/engine_authsession.go

@@ -0,0 +1,108 @@
+package internal
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+	"google.golang.org/protobuf/types/known/timestamppb"
+
+	"github.com/netbirdio/netbird/client/internal/auth/sessionwatch"
+	cProto "github.com/netbirdio/netbird/client/proto"
+	"github.com/netbirdio/netbird/client/system"
+)
+
+// ApplySessionDeadline propagates the absolute SSO session deadline carried on
+// LoginResponse / SyncResponse to both the watcher (for the edge-triggered
+// warning) and the status recorder (for the SubscribeStatus / Status RPC
+// snapshot the UI consumes).
+//
+// The wire field is 3-state:
+//   - nil                        → snapshot carries no info; keep the
+//     previously-anchored deadline (no-op)
+//   - explicit zero (s=0, n=0)   → peer is not SSO-registered or expiry is
+//     disabled; clear both sinks
+//   - valid timestamp            → new deadline; arm watcher, expose on
+//     status recorder
+//
+// Deadline sanity-checks live in sessionwatch.Watcher.Update. Any rejected
+// value is treated as a clear on both sinks: the alternative — leaving the
+// previously-known deadline in place — risks the UI confidently displaying
+// a stale "expires in X" while the server has actually invalidated it.
+func (e *Engine) ApplySessionDeadline(ts *timestamppb.Timestamp) {
+	if ts == nil {
+		return
+	}
+	var deadline time.Time
+	// Explicit zero (seconds=0 AND nanos=0) is the sentinel for "disabled".
+	// Everything else flows through Watcher.Update, whose sanity-checks
+	// reject out-of-range / pre-epoch / far-future / too-stale values and
+	// clear on rejection.
+	if ts.GetSeconds() != 0 || ts.GetNanos() != 0 {
+		deadline = ts.AsTime().UTC()
+	}
+	if e.sessionWatcher == nil {
+		return
+	}
+	// Watcher.Update owns the propagation to the status recorder (the
+	// SubscribeStatus / Status snapshot the UI reads): a set writes the
+	// deadline, a clear or a sanity-check rejection writes the zero value.
+	// Keeping a single writer is what stops the recorder from drifting out
+	// of sync with the warning timers.
+	if err := e.sessionWatcher.Update(deadline); err != nil {
+		log.Errorf("auth session deadline rejected: %v, clearing", err)
+		e.statusRecorder.PublishEvent(
+			cProto.SystemEvent_ERROR,
+			cProto.SystemEvent_AUTHENTICATION,
+			"session deadline rejected",
+			"",
+			map[string]string{sessionwatch.MetaSessionDeadlineRejected: err.Error()},
+		)
+	}
+}
+
+// DismissSessionWarning records the user's "Dismiss" click on the
+// T-WarningLead interactive notification and suppresses the upcoming
+// T-FinalWarningLead fallback for the current deadline. No-op when the
+// watcher is not running or holds no deadline.
+func (e *Engine) DismissSessionWarning() {
+	if e.sessionWatcher == nil {
+		return
+	}
+	e.sessionWatcher.Dismiss()
+}
+
+// ExtendAuthSession asks the management server to refresh the SSO session
+// expiry deadline using the supplied JWT, then mirrors the new deadline into
+// the daemon's state. The tunnel is untouched; no resync, no reconnect.
+//
+// Returns the new absolute UTC deadline (or zero time when the server
+// reports the peer is not eligible for extension).
+func (e *Engine) ExtendAuthSession(ctx context.Context, jwtToken string) (time.Time, error) {
+	if jwtToken == "" {
+		return time.Time{}, errors.New("jwt token is required")
+	}
+	if e.mgmClient == nil {
+		return time.Time{}, errors.New("management client is not initialised")
+	}
+
+	info, err := system.GetInfoWithChecks(ctx, e.checks)
+	if err != nil {
+		log.Warnf("failed to collect system info for session extend: %v", err)
+		info = system.GetInfo(ctx)
+	}
+
+	resp, err := e.mgmClient.ExtendAuthSession(info, jwtToken)
+	if err != nil {
+		return time.Time{}, fmt.Errorf("extend auth session on management: %w", err)
+	}
+
+	e.ApplySessionDeadline(resp.GetSessionExpiresAt())
+
+	if resp.GetSessionExpiresAt().IsValid() {
+		return resp.GetSessionExpiresAt().AsTime().UTC(), nil
+	}
+	return time.Time{}, nil
+}

client/internal/engine_privileged_test.go

@@ -124,7 +124,7 @@ func TestEngine_SSH(t *testing.T) {
 		RemotePeersIsEmpty: false,
 	}
 
-	_, err = engine.updateNetworkMap(networkMap, maxPeersPerSyncPass, true)
+	err = engine.updateNetworkMap(networkMap)
 	require.NoError(t, err)
 
 	assert.Nil(t, engine.sshServer)
@@ -146,7 +146,7 @@ func TestEngine_SSH(t *testing.T) {
 		RemotePeersIsEmpty: false,
 	}
 
-	_, err = engine.updateNetworkMap(networkMap, maxPeersPerSyncPass, true)
+	err = engine.updateNetworkMap(networkMap)
 	require.NoError(t, err)
 
 	time.Sleep(250 * time.Millisecond)
@@ -159,7 +159,7 @@ func TestEngine_SSH(t *testing.T) {
 		RemotePeersIsEmpty: false,
 	}
 
-	_, err = engine.updateNetworkMap(networkMap, maxPeersPerSyncPass, true)
+	err = engine.updateNetworkMap(networkMap)
 	require.NoError(t, err)
 
 	// time.Sleep(250 * time.Millisecond)
@@ -174,7 +174,7 @@ func TestEngine_SSH(t *testing.T) {
 		RemotePeersIsEmpty: false,
 	}
 
-	_, err = engine.updateNetworkMap(networkMap, maxPeersPerSyncPass, true)
+	err = engine.updateNetworkMap(networkMap)
 	require.NoError(t, err)
 
 	assert.Nil(t, engine.sshServer)

client/internal/engine_session_deadline_test.go

@@ -0,0 +1,78 @@
+package internal
+
+import (
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+	"google.golang.org/protobuf/types/known/timestamppb"
+
+	"github.com/netbirdio/netbird/client/internal/auth/sessionwatch"
+	"github.com/netbirdio/netbird/client/internal/peer"
+)
+
+// TestApplySessionDeadline_ThreeState pins down the 3-state semantics of the
+// wire field carried on LoginResponse / SyncResponse:
+//
+//   - nil pointer            → no info; previously-anchored deadline survives
+//   - explicit zero value    → "expiry disabled" sentinel; both sinks cleared
+//   - valid future timestamp → new deadline propagated to both sinks
+func TestApplySessionDeadline_ThreeState(t *testing.T) {
+	newEngine := func() *Engine {
+		recorder := peer.NewRecorder("")
+		return &Engine{
+			statusRecorder: recorder,
+			sessionWatcher: sessionwatch.New(recorder),
+		}
+	}
+
+	t.Run("valid timestamp sets deadline on both sinks", func(t *testing.T) {
+		e := newEngine()
+		deadline := time.Now().Add(time.Hour).UTC().Truncate(time.Second)
+
+		e.ApplySessionDeadline(timestamppb.New(deadline))
+
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(deadline),
+			"status recorder should hold the new deadline")
+	})
+
+	t.Run("nil is a no-op and preserves previous deadline", func(t *testing.T) {
+		e := newEngine()
+		seeded := time.Now().Add(time.Hour).UTC().Truncate(time.Second)
+		e.ApplySessionDeadline(timestamppb.New(seeded))
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(seeded))
+
+		e.ApplySessionDeadline(nil)
+
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(seeded),
+			"nil snapshot must not disturb the existing deadline")
+	})
+
+	t.Run("explicit zero clears a previously-anchored deadline", func(t *testing.T) {
+		e := newEngine()
+		seeded := time.Now().Add(time.Hour).UTC().Truncate(time.Second)
+		e.ApplySessionDeadline(timestamppb.New(seeded))
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(seeded))
+
+		// Explicit zero Timestamp{} (seconds=0, nanos=0) is the
+		// "expiry disabled / not SSO" sentinel.
+		e.ApplySessionDeadline(&timestamppb.Timestamp{})
+
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().IsZero(),
+			"explicit zero sentinel must clear the deadline")
+	})
+
+	t.Run("invalid timestamp clears the deadline", func(t *testing.T) {
+		e := newEngine()
+		seeded := time.Now().Add(time.Hour).UTC().Truncate(time.Second)
+		e.ApplySessionDeadline(timestamppb.New(seeded))
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(seeded))
+
+		// Out-of-range nanos → IsValid()==false; same-meaning as the
+		// disabled sentinel for downstream sinks.
+		e.ApplySessionDeadline(&timestamppb.Timestamp{Seconds: 1, Nanos: -1})
+
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().IsZero(),
+			"invalid timestamp must clear the deadline")
+	})
+}

client/internal/engine_sessionwatch.go

@@ -0,0 +1,16 @@
+//go:build !js
+
+package internal
+
+import (
+	"github.com/netbirdio/netbird/client/internal/auth/sessionwatch"
+	"github.com/netbirdio/netbird/client/internal/peer"
+)
+
+// newSessionWatcher returns the real SSO session expiry watcher for every
+// non-wasm build. The js/wasm build gets a no-op stub from
+// engine_sessionwatch_js.go so the sessionwatch package (and its timer
+// machinery) never links into the wasm binary.
+func newSessionWatcher(recorder *peer.Status) sessionDeadlineWatcher {
+	return sessionwatch.New(recorder)
+}

client/internal/engine_sessionwatch_js.go

@@ -0,0 +1,44 @@
+//go:build js
+
+package internal
+
+import (
+	"time"
+
+	"github.com/netbirdio/netbird/client/internal/peer"
+)
+
+// noopSessionWatcher is the js/wasm stand-in for sessionwatch.Watcher. The
+// wasm client never runs the engine's session-warning flow (the interactive
+// T-WarningLead notification and the T-FinalWarningLead fallback dialog live
+// in the desktop UI), so linking the full sessionwatch package (timers, event
+// composition) would only bloat the binary.
+//
+// It still mirrors the deadline into the status recorder so the SubscribeStatus
+// / Status snapshot the UI consumes stays correct — only the timer-driven
+// warnings are dropped.
+type noopSessionWatcher struct {
+	recorder *peer.Status
+}
+
+func newSessionWatcher(recorder *peer.Status) sessionDeadlineWatcher {
+	return noopSessionWatcher{recorder: recorder}
+}
+
+// Update mirrors the real watcher's recorder propagation without the timers or
+// sanity-check sentinels: a valid deadline is exposed on the status snapshot,
+// the zero time clears it.
+func (w noopSessionWatcher) Update(deadline time.Time) error {
+	if w.recorder != nil {
+		w.recorder.SetSessionExpiresAt(deadline)
+	}
+	return nil
+}
+
+func (noopSessionWatcher) Dismiss() {
+	// No-op: only suppresses the timer-driven final-warning, which this stub never arms.
+}
+
+func (noopSessionWatcher) Close() {
+	// No-op: no timers to stop and no state to unwind; the recorder is cleared via Update(zero).
+}

client/internal/engine_test.go

@@ -437,7 +437,7 @@ func TestEngine_UpdateNetworkMap(t *testing.T) {
 
 	for _, c := range []testCase{case1, case2, case3, case4, case5, case6} {
 		t.Run(c.name, func(t *testing.T) {
-			_, err = engine.updateNetworkMap(c.networkMap, maxPeersPerSyncPass, true)
+			err = engine.updateNetworkMap(c.networkMap)
 			if err != nil {
 				t.Fatal(err)
 				return
@@ -464,47 +464,6 @@ func TestEngine_UpdateNetworkMap(t *testing.T) {
 			}
 		})
 	}
-
-	// chunked apply: with a per-pass cap smaller than the number of peers, a
-	// single updateNetworkMap applies one batch and reports more==true; the
-	// caller re-runs until convergence. (engine currently holds 0 peers.)
-	t.Run("chunked add converges over multiple passes", func(t *testing.T) {
-		nm := &mgmtProto.NetworkMap{
-			Serial:      6,
-			RemotePeers: []*mgmtProto.RemotePeerConfig{peer1, peer2, peer3},
-		}
-
-		more, err := engine.updateNetworkMap(nm, 1, true)
-		require.NoError(t, err)
-		require.True(t, more, "pass 1 should signal more")
-		require.Len(t, engine.peerStore.PeersPubKey(), 1)
-
-		more, err = engine.updateNetworkMap(nm, 1, false)
-		require.NoError(t, err)
-		require.True(t, more, "pass 2 should signal more")
-		require.Len(t, engine.peerStore.PeersPubKey(), 2)
-
-		more, err = engine.updateNetworkMap(nm, 1, false)
-		require.NoError(t, err)
-		require.False(t, more, "pass 3 should converge")
-		require.Len(t, engine.peerStore.PeersPubKey(), 3)
-	})
-
-	t.Run("chunked remove converges over multiple passes", func(t *testing.T) {
-		nm := &mgmtProto.NetworkMap{
-			Serial:      7,
-			RemotePeers: []*mgmtProto.RemotePeerConfig{peer1}, // remove peer2, peer3
-		}
-
-		more, err := engine.updateNetworkMap(nm, 1, true)
-		require.NoError(t, err)
-		require.True(t, more, "pass 1 should signal more (2 to remove, cap 1)")
-
-		more, err = engine.updateNetworkMap(nm, 1, false)
-		require.NoError(t, err)
-		require.False(t, more, "pass 2 should converge")
-		require.Len(t, engine.peerStore.PeersPubKey(), 1)
-	})
 }
 
 func TestEngine_UpdateNetworkMapWithRoutes(t *testing.T) {
@@ -675,7 +634,7 @@ func TestEngine_UpdateNetworkMapWithRoutes(t *testing.T) {
 				}
 			}()
 
-			_, err = engine.updateNetworkMap(testCase.networkMap, maxPeersPerSyncPass, true)
+			err = engine.updateNetworkMap(testCase.networkMap)
 			assert.NoError(t, err, "shouldn't return error")
 			assert.Equal(t, testCase.expectedSerial, input.inputSerial, "serial should match")
 			assert.Len(t, input.clientRoutes, testCase.expectedLen, "clientRoutes len should match")
@@ -879,7 +838,7 @@ func TestEngine_UpdateNetworkMapWithDNSUpdate(t *testing.T) {
 				}
 			}()
 
-			_, err = engine.updateNetworkMap(testCase.networkMap, maxPeersPerSyncPass, true)
+			err = engine.updateNetworkMap(testCase.networkMap)
 			assert.NoError(t, err, "shouldn't return error")
 			assert.Equal(t, testCase.expectedSerial, input.inputSerial, "serial should match")
 			assert.Len(t, input.inputNSGroups, testCase.expectedZonesLen, "zones len should match")

client/internal/mapsync.go

@@ -1,214 +0,0 @@
-package internal
-
-import (
-	"context"
-	"sync"
-	"time"
-
-	log "github.com/sirupsen/logrus"
-
-	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
-)
-
-// mapStateManager is the single read/write point between the management stream
-// (writes) and the convergence loop (reads/applies).
-//
-// The stream calls SetTarget with the latest full SyncResponse — the complete
-// desired state. A single background goroutine (run) applies it to the engine in
-// bounded passes via apply() until converged, releasing syncMsgMux between passes
-// so other subsystems interleave. If a newer update arrives mid-flight, the loop
-// coalesces: it keeps converging toward the latest target and the intermediate one
-// is SKIPPED — never applied on its own (logged, no onConverged).
-//
-// Convergence is a single comparison: appliedGen == targetGen. targetGen
-// increments on every SetTarget (an internal generation counter, so it also covers
-// config-only updates that carry no network-map serial).
-//
-// onConverged fires once for each — and only each — map that is actually processed
-// (i.e. converged as the target). Skipped/superseded maps and dropped-on-error maps
-// do NOT fire it. So "sync finished in X" / RecordSyncDuration always corresponds
-// to a real, completed alignment.
-type mapStateManager struct {
-	// apply performs one bounded apply pass and reports whether more passes are needed.
-	// firstPass is true on the first pass of a given target, so the caller can run
-	// wholesale (firewall/routes/DNS/forward-rules) once per target and skip it on the
-	// re-runs that only drain the bounded peer batches. The manager owns this signal
-	// because it owns the convergence boundary; the engine need not track serials for it.
-	apply func(update *mgmProto.SyncResponse, firstPass bool) (bool, error)
-	// onConverged is called once per processed map, with the elapsed time since that
-	// map was received (for the sync-duration metric / "sync finished" log).
-	onConverged func(time.Duration)
-	// persist snapshots an update to disk for restore-on-restart. Called once per
-	// update received from management (in SetTarget), including ones later coalesced
-	// or skipped from apply, so the on-disk state mirrors what management last sent.
-	// The impl skips config-only updates (nil NetworkMap). May be nil.
-	persist func(*mgmProto.SyncResponse)
-
-	mu          sync.Mutex
-	target      *mgmProto.SyncResponse
-	targetGen   uint64
-	appliedGen  uint64
-	targetSetAt time.Time
-
-	wake chan struct{}
-}
-
-func newMapStateManager(apply func(update *mgmProto.SyncResponse, firstPass bool) (bool, error), persist func(*mgmProto.SyncResponse), onConverged func(time.Duration)) *mapStateManager {
-	return &mapStateManager{
-		apply:       apply,
-		persist:     persist,
-		onConverged: onConverged,
-		wake:        make(chan struct{}, 1),
-	}
-}
-
-// SetTarget records the latest update as the desired state and wakes the loop.
-// It returns immediately; convergence happens in the background. Serial-based
-// staleness of the network map is still enforced inside apply (updateNetworkMap).
-func (m *mapStateManager) SetTarget(update *mgmProto.SyncResponse) error {
-	m.mu.Lock()
-	// A target that has not settled yet (targetGen > appliedGen) is being superseded
-	// before it converged: we coalesce to the latest map and never apply this one on
-	// its own. It is SKIPPED — logged here, and it will not fire onConverged.
-	if m.target != nil && m.targetGen > m.appliedGen {
-		log.Debugf("sync map (gen %d) superseded before convergence, skipping", m.targetGen)
-	}
-	m.target = m.mergeTarget(m.target, update)
-	// Bump an internal generation counter, NOT the map serial: config-only updates
-	// (relay token rotation, STUN/TURN) arrive with NetworkMap == nil and carry no
-	// serial, yet must still be applied. Every SetTarget is therefore a distinct
-	// target regardless of payload. Map-serial staleness is enforced separately
-	// inside apply (updateNetworkMap).
-	m.targetGen++
-	m.targetSetAt = time.Now()
-	m.mu.Unlock()
-
-	select {
-	case m.wake <- struct{}{}:
-	default:
-	}
-
-	// Persist every update received from management — once per update (not per apply
-	// pass), and including ones that get coalesced/skipped from apply, so the on-disk
-	// state always reflects the latest map management sent. Done after waking the loop
-	// so convergence can start in parallel with the disk write. The persist impl skips
-	// config-only updates (nil NetworkMap).
-	if m.persist != nil {
-		m.persist(update)
-	}
-	return nil
-}
-
-// mergeTarget combines the currently pending target with a freshly received update
-// and returns the new desired state. It is called under m.mu from SetTarget and is
-// the single seam where the replace-vs-squash decision lives.
-//
-// Today management always sends a FULL map (the complete desired state), so the
-// update simply replaces whatever was pending — prev is ignored. When management
-// starts sending incremental/delta updates, squash `update` onto `prev` here; the
-// rest of the manager (generation tracking, convergence, signaling) is unaffected
-// because it already treats target as "the complete desired state, whatever it is".
-func (m *mapStateManager) mergeTarget(prev, update *mgmProto.SyncResponse) *mgmProto.SyncResponse {
-	// Nothing pending to preserve (no prev, or prev already fully applied): plain replace.
-	if prev == nil || update == nil || m.targetGen == m.appliedGen {
-		return update
-	}
-
-	// prev still has unapplied state (targetGen > appliedGen). In the sync protocol a
-	// nil component means "no change", so if `update` omits a component that prev
-	// carried, carry prev's forward — otherwise coalescing an update that superseded a
-	// not-yet-applied one would silently drop the map or config it uniquely brought.
-	// A present component in `update` is newer and wins. Management may send map-only
-	// updates (nil config) and config-only updates (nil map); both are handled here.
-	// A nil component in `update` means "no change", so fill it in from prev — otherwise
-	// coalescing an update that superseded a not-yet-applied one would drop the map or
-	// config it uniquely carried. A present component in `update` is newer and wins.
-	// We mutate `update` in place: it is a fresh per-message allocation from the sync
-	// stream (see receiveUpdatesEvents — not reused), and persisting this squashed target
-	// is correct, since it is the current full (superset) desired state.
-	if update.GetNetworkMap() == nil && prev.GetNetworkMap() != nil {
-		update.NetworkMap = prev.GetNetworkMap()
-		update.Checks = prev.Checks // checks travel with the map
-	}
-	if update.GetNetbirdConfig() == nil && prev.GetNetbirdConfig() != nil {
-		update.NetbirdConfig = prev.GetNetbirdConfig()
-	}
-	return update
-}
-
-// run drives convergence until ctx is done. It is meant to run in its own goroutine.
-func (m *mapStateManager) run(ctx context.Context) {
-	// passGen is the generation of the most recent apply() call (0 = none). A pass is
-	// the first for its target when its generation differs from the previous one —
-	// true on a fresh target and on a coalesced switch to a newer target mid-flight.
-	var passGen uint64
-	for {
-		m.mu.Lock()
-		target, tg, ag := m.target, m.targetGen, m.appliedGen
-		m.mu.Unlock()
-
-		// Fully converged (or nothing yet): block until a new target arrives.
-		if target == nil || ag == tg {
-			select {
-			case <-ctx.Done():
-				return
-			case <-m.wake:
-				continue
-			}
-		}
-
-		firstPass := tg != passGen
-		passGen = tg
-		more, err := m.apply(target, firstPass)
-		if err != nil {
-			if ctx.Err() != nil {
-				return
-			}
-			// Log and DROP this target — do not retry it. A deterministic failure
-			// (e.g. a malformed peer in the map) would otherwise spin every pass
-			// making no progress. Management is the source of truth and re-delivers
-			// the full map on the next sync, so dropping is safe; peers already
-			// applied this convergence stay (idempotent diffs) and the remainder is
-			// reconciled by the next target. Mirrors the legacy handleSync path,
-			// where the apply error was logged by the gRPC client and the update
-			// dropped. No onConverged: this target did not converge.
-			log.Errorf("apply sync pass, dropping update: %v", err)
-			m.settle(tg, false)
-			continue
-		}
-
-		if more {
-			// keep converging the current target; syncMsgMux was released by apply
-			// between passes so other subsystems interleave.
-			continue
-		}
-
-		// This pass converged. Mark applied and signal this one map.
-		m.settle(tg, true)
-		// if a newer target arrived mid-pass, settle is a no-op (targetGen != tg) and
-		// ag<tg next iteration -> apply it; this generation was skipped (logged in
-		// SetTarget) and is not signaled.
-	}
-}
-
-// settle marks generation tg as processed so the loop goes idle instead of
-// re-applying the same target. It is a no-op when a newer target arrived during the
-// pass (targetGen != tg), leaving appliedGen behind so that target re-applies — the
-// just-finished generation was already counted as skipped.
-//
-// When signal is true (the pass converged) it fires onConverged once for this map;
-// when false (the target was dropped on error) it does not — the map did not converge.
-func (m *mapStateManager) settle(tg uint64, signal bool) {
-	m.mu.Lock()
-	if m.targetGen != tg {
-		m.mu.Unlock()
-		return
-	}
-	m.appliedGen = tg
-	setAt := m.targetSetAt
-	m.mu.Unlock()
-
-	if signal && m.onConverged != nil {
-		m.onConverged(time.Since(setAt))
-	}
-}

client/internal/mapsync_test.go

@@ -1,281 +0,0 @@
-package internal
-
-import (
-	"context"
-	"errors"
-	"sync/atomic"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/require"
-
-	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
-)
-
-// mergeTarget fills components missing from the incoming update with the pending
-// (not-yet-applied) prev's, in place, so a coalesced/superseded update does not drop
-// the map or config it uniquely carried.
-func TestMapStateManager_MergeTargetPreservesPendingState(t *testing.T) {
-	m := newMapStateManager(nil, nil, nil)
-
-	// config-only update while a full map is still converging (targetGen > appliedGen):
-	// the pending map (+ checks) is filled into the update in place
-	m.targetGen, m.appliedGen = 5, 4
-	prev := &mgmProto.SyncResponse{NetworkMap: &mgmProto.NetworkMap{Serial: 5}}
-	update := &mgmProto.SyncResponse{NetbirdConfig: &mgmProto.NetbirdConfig{}}
-	merged := m.mergeTarget(prev, update)
-	require.Same(t, update, merged, "merges in place, returns the update")
-	require.EqualValues(t, 5, merged.GetNetworkMap().GetSerial(), "pending map preserved")
-	require.NotNil(t, merged.GetNetbirdConfig(), "new config kept")
-
-	// symmetric: map-only update while a config-only update is pending -> keep the config
-	m.targetGen, m.appliedGen = 5, 4
-	prev = &mgmProto.SyncResponse{NetbirdConfig: &mgmProto.NetbirdConfig{}}
-	update = &mgmProto.SyncResponse{NetworkMap: &mgmProto.NetworkMap{Serial: 7}}
-	merged = m.mergeTarget(prev, update)
-	require.EqualValues(t, 7, merged.GetNetworkMap().GetSerial(), "new map kept")
-	require.NotNil(t, merged.GetNetbirdConfig(), "pending config preserved")
-
-	// prev already applied (targetGen == appliedGen): plain replace, no fill-in
-	m.targetGen, m.appliedGen = 5, 5
-	prev = &mgmProto.SyncResponse{NetworkMap: &mgmProto.NetworkMap{Serial: 5}}
-	update = &mgmProto.SyncResponse{NetbirdConfig: &mgmProto.NetbirdConfig{}}
-	merged = m.mergeTarget(prev, update)
-	require.Same(t, update, merged)
-	require.Nil(t, merged.GetNetworkMap(), "no map grafted when prev already applied")
-
-	// nothing to carry (update has a map, prev has no config): plain replace
-	m.targetGen, m.appliedGen = 5, 4
-	prev = &mgmProto.SyncResponse{NetworkMap: &mgmProto.NetworkMap{Serial: 5}}
-	update = &mgmProto.SyncResponse{NetworkMap: &mgmProto.NetworkMap{Serial: 6}}
-	require.Same(t, update, m.mergeTarget(prev, update))
-}
-
-// converges over the bounded passes (apply returns more until the 3rd pass),
-// fires onConverged exactly once, then blocks (no further apply) until a new target.
-func TestMapStateManager_ConvergesThenStops(t *testing.T) {
-	var passes int32
-	var firstPasses int32
-	converged := make(chan struct{}, 1)
-
-	apply := func(_ *mgmProto.SyncResponse, firstPass bool) (bool, error) {
-		n := atomic.AddInt32(&passes, 1)
-		if firstPass {
-			atomic.AddInt32(&firstPasses, 1)
-		}
-		return n < 3, nil // more on pass 1 and 2, converge on pass 3
-	}
-	m := newMapStateManager(apply, nil, func(time.Duration) { converged <- struct{}{} })
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-
-	select {
-	case <-converged:
-	case <-time.After(2 * time.Second):
-		t.Fatal("manager did not converge")
-	}
-	require.EqualValues(t, 3, atomic.LoadInt32(&passes))
-	require.EqualValues(t, 1, atomic.LoadInt32(&firstPasses), "firstPass true only on pass 1, false on re-runs of the same target")
-
-	// once converged the loop blocks: no further apply calls
-	time.Sleep(100 * time.Millisecond)
-	require.EqualValues(t, 3, atomic.LoadInt32(&passes), "apply must not run after convergence")
-}
-
-// persist runs once per received update (not per apply pass), regardless of how many
-// bounded passes that target takes to converge.
-func TestMapStateManager_PersistsOncePerUpdate(t *testing.T) {
-	var passes, persists int32
-	converged := make(chan struct{}, 1)
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		n := atomic.AddInt32(&passes, 1)
-		return n < 3, nil // 3 passes for one target
-	}
-	persist := func(*mgmProto.SyncResponse) { atomic.AddInt32(&persists, 1) }
-	m := newMapStateManager(apply, persist, func(time.Duration) { converged <- struct{}{} })
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	select {
-	case <-converged:
-	case <-time.After(2 * time.Second):
-		t.Fatal("did not converge")
-	}
-	require.EqualValues(t, 3, atomic.LoadInt32(&passes))
-	require.EqualValues(t, 1, atomic.LoadInt32(&persists), "persist once per update, not per pass")
-}
-
-// every update received from management is persisted — even one that is coalesced /
-// skipped from apply before it ever converges.
-func TestMapStateManager_PersistsEveryUpdateIncludingSkipped(t *testing.T) {
-	release := make(chan struct{})
-	var persists int32
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		<-release // hold the first apply so the second update coalesces/skips
-		return false, nil
-	}
-	persist := func(*mgmProto.SyncResponse) { atomic.AddInt32(&persists, 1) }
-	m := newMapStateManager(apply, persist, nil)
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{})) // map1 -> apply blocks
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{})) // map2 supersedes map1 (skipped from apply)
-	close(release)
-
-	// both updates persisted even though map1 is skipped from apply
-	require.Eventually(t, func() bool { return atomic.LoadInt32(&persists) == 2 }, 2*time.Second, 10*time.Millisecond)
-}
-
-// each map that is actually processed (converged before the next arrives) fires
-// onConverged exactly once — mirroring the legacy per-message handleSync timing.
-func TestMapStateManager_SignalsEachProcessedMap(t *testing.T) {
-	converged := make(chan struct{}, 8)
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		return false, nil // converge in one pass
-	}
-	m := newMapStateManager(apply, nil, func(time.Duration) { converged <- struct{}{} })
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	const maps = 3
-	for i := 0; i < maps; i++ {
-		require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-		select { // wait for this map to converge before sending the next (no coalescing)
-		case <-converged:
-		case <-time.After(2 * time.Second):
-			t.Fatalf("map %d not signaled", i)
-		}
-	}
-
-	// no extra signals once the stream goes quiet
-	select {
-	case <-converged:
-		t.Fatal("unexpected extra onConverged")
-	case <-time.After(100 * time.Millisecond):
-	}
-}
-
-// a map superseded before it converges is skipped: only the latest (processed) map
-// fires onConverged, not the skipped one.
-func TestMapStateManager_SkippedMapNotSignaled(t *testing.T) {
-	release := make(chan struct{})
-	var applies, converged atomic.Int32
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		applies.Add(1)
-		<-release // hold the first apply in-flight so we can queue a newer target
-		return false, nil
-	}
-	m := newMapStateManager(apply, nil, func(time.Duration) { converged.Add(1) })
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	// map1 is picked up; its apply blocks on release
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	require.Eventually(t, func() bool { return applies.Load() >= 1 }, 2*time.Second, 5*time.Millisecond)
-
-	// map2 supersedes map1 before it settled -> map1 is skipped
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	close(release) // let both applies proceed
-
-	// only the processed (latest) map signals; the skipped one does not
-	require.Eventually(t, func() bool { return converged.Load() == 1 }, 2*time.Second, 10*time.Millisecond)
-	time.Sleep(150 * time.Millisecond)
-	require.EqualValues(t, 1, converged.Load(), "skipped map must not fire onConverged")
-	require.EqualValues(t, 2, applies.Load(), "both targets entered apply (map1 once, map2 once)")
-}
-
-// an apply error drops the target: no retry of the same target, no onConverged,
-// the loop goes idle — and a fresh target is still applied afterwards.
-func TestMapStateManager_DropsTargetOnError(t *testing.T) {
-	applied := make(chan struct{}, 8)
-	var failNext atomic.Bool
-	failNext.Store(true)
-
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		applied <- struct{}{}
-		if failNext.Load() {
-			return false, errors.New("boom")
-		}
-		return false, nil // converge in one pass
-	}
-	var converged atomic.Int32
-	m := newMapStateManager(apply, nil, func(time.Duration) { converged.Add(1) })
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	// first target errors -> applied once, then dropped (no retry, no onConverged)
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	select {
-	case <-applied:
-	case <-time.After(2 * time.Second):
-		t.Fatal("errored target not applied")
-	}
-	select {
-	case <-applied:
-		t.Fatal("errored target must not be retried")
-	case <-time.After(150 * time.Millisecond):
-	}
-	require.EqualValues(t, 0, converged.Load(), "onConverged must not fire on error")
-
-	// a new target is still processed normally and converges
-	failNext.Store(false)
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	select {
-	case <-applied:
-	case <-time.After(2 * time.Second):
-		t.Fatal("new target after error not applied")
-	}
-	require.Eventually(t, func() bool { return converged.Load() == 1 }, 2*time.Second, 10*time.Millisecond)
-}
-
-// a new target after convergence triggers a fresh apply; an idle (converged)
-// manager does not apply on its own.
-func TestMapStateManager_ReappliesOnNewTarget(t *testing.T) {
-	applied := make(chan struct{}, 8)
-	apply := func(_ *mgmProto.SyncResponse, _ bool) (bool, error) {
-		applied <- struct{}{}
-		return false, nil // converge in one pass
-	}
-	m := newMapStateManager(apply, nil, nil)
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-	go m.run(ctx)
-
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	select {
-	case <-applied:
-	case <-time.After(2 * time.Second):
-		t.Fatal("first target not applied")
-	}
-
-	// converged → must stay idle (no spurious apply)
-	select {
-	case <-applied:
-		t.Fatal("unexpected apply while idle/converged")
-	case <-time.After(150 * time.Millisecond):
-	}
-
-	require.NoError(t, m.SetTarget(&mgmProto.SyncResponse{}))
-	select {
-	case <-applied:
-	case <-time.After(2 * time.Second):
-		t.Fatal("new target not applied")
-	}
-}

client/internal/netflow/logger/logger.go

@@ -27,7 +27,7 @@ type Logger struct {
 	wgIfaceNetV6       netip.Prefix
 	dnsCollection      atomic.Bool
 	exitNodeCollection atomic.Bool
-	Store              types.Store
+	Store              types.AggregatingStore
 }
 
 func New(statusRecorder *peer.Status, wgIfaceIPNet, wgIfaceIPNetV6 netip.Prefix) *Logger {
@@ -35,7 +35,7 @@ func New(statusRecorder *peer.Status, wgIfaceIPNet, wgIfaceIPNetV6 netip.Prefix)
 		statusRecorder: statusRecorder,
 		wgIfaceNet:     wgIfaceIPNet,
 		wgIfaceNetV6:   wgIfaceIPNetV6,
-		Store:          store.NewMemoryStore(),
+		Store:          store.NewAggregatingMemoryStore(),
 	}
 }
 
@@ -125,6 +125,10 @@ func (l *Logger) stop() {
 	l.mux.Unlock()
 }
 
+func (l *Logger) ResetAggregationWindow() types.FlowEventAggregator {
+	return l.Store.ResetAggregationWindow()
+}
+
 func (l *Logger) GetEvents() []*types.Event {
 	return l.Store.GetEvents()
 }

client/internal/netflow/manager.go

@@ -9,12 +9,14 @@ import (
 	"sync"
 	"time"
 
+	"github.com/cenkalti/backoff/v4"
 	"github.com/google/uuid"
 	log "github.com/sirupsen/logrus"
 	"google.golang.org/protobuf/types/known/timestamppb"
 
 	"github.com/netbirdio/netbird/client/internal/netflow/conntrack"
 	"github.com/netbirdio/netbird/client/internal/netflow/logger"
+	"github.com/netbirdio/netbird/client/internal/netflow/store"
 	nftypes "github.com/netbirdio/netbird/client/internal/netflow/types"
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/flow/client"
@@ -23,14 +25,16 @@ import (
 
 // Manager handles netflow tracking and logging
 type Manager struct {
-	mux            sync.Mutex
-	shutdownWg     sync.WaitGroup
-	logger         nftypes.FlowLogger
-	flowConfig     *nftypes.FlowConfig
-	conntrack      nftypes.ConnTracker
-	receiverClient *client.GRPCClient
-	publicKey      []byte
-	cancel         context.CancelFunc
+	mux               sync.Mutex
+	shutdownWg        sync.WaitGroup
+	logger            nftypes.FlowLogger
+	flowConfig        *nftypes.FlowConfig
+	conntrack         nftypes.ConnTracker
+	receiverClient    *client.GRPCClient
+	eventsWithoutAcks nftypes.Store
+	publicKey         []byte
+	cancel            context.CancelFunc
+	retryInterval     time.Duration
 }
 
 // NewManager creates a new netflow manager
@@ -48,9 +52,11 @@ func NewManager(iface nftypes.IFaceMapper, publicKey []byte, statusRecorder *pee
 	}
 
 	return &Manager{
-		logger:    flowLogger,
-		conntrack: ct,
-		publicKey: publicKey,
+		logger:            flowLogger,
+		conntrack:         ct,
+		publicKey:         publicKey,
+		retryInterval:     time.Second,
+		eventsWithoutAcks: store.NewMemoryStore(),
 	}
 }
 
@@ -66,6 +72,7 @@ func (m *Manager) needsNewClient(previous *nftypes.FlowConfig) bool {
 }
 
 // enableFlow starts components for flow tracking
+// must be called under m.mux lock
 func (m *Manager) enableFlow(previous *nftypes.FlowConfig) error {
 	// first make sender ready so events don't pile up
 	if m.needsNewClient(previous) {
@@ -85,6 +92,7 @@ func (m *Manager) enableFlow(previous *nftypes.FlowConfig) error {
 	return nil
 }
 
+// must be called under m.mux lock
 func (m *Manager) resetClient() error {
 	if m.receiverClient != nil {
 		if err := m.receiverClient.Close(); err != nil {
@@ -107,14 +115,19 @@ func (m *Manager) resetClient() error {
 	ctx, cancel := context.WithCancel(context.Background())
 	m.cancel = cancel
 
-	m.shutdownWg.Add(2)
+	m.shutdownWg.Add(3)
+	flowConfigInterval := m.flowConfig.Interval
 	go func() {
 		defer m.shutdownWg.Done()
-		m.receiveACKs(ctx, flowClient)
+		m.receiveACKs(ctx, flowClient, flowConfigInterval)
 	}()
 	go func() {
 		defer m.shutdownWg.Done()
-		m.startSender(ctx)
+		m.startSender(ctx, flowConfigInterval)
+	}()
+	go func() {
+		defer m.shutdownWg.Done()
+		m.startRetries(ctx, flowConfigInterval)
 	}()
 
 	return nil
@@ -198,8 +211,8 @@ func (m *Manager) GetLogger() nftypes.FlowLogger {
 	return m.logger
 }
 
-func (m *Manager) startSender(ctx context.Context) {
-	ticker := time.NewTicker(m.flowConfig.Interval)
+func (m *Manager) startSender(ctx context.Context, flowConfigInterval time.Duration) {
+	ticker := time.NewTicker(flowConfigInterval)
 	defer ticker.Stop()
 
 	for {
@@ -207,27 +220,29 @@ func (m *Manager) startSender(ctx context.Context) {
 		case <-ctx.Done():
 			return
 		case <-ticker.C:
-			events := m.logger.GetEvents()
+			collectedEvents := m.logger.ResetAggregationWindow()
+			events := collectedEvents.GetAggregatedEvents()
 			for _, event := range events {
+				m.eventsWithoutAcks.StoreEvent(event)
 				if err := m.send(event); err != nil {
 					log.Errorf("failed to send flow event to server: %v", err)
-					continue
+				} else {
+					log.Tracef("sent flow event: %s", event.ID)
 				}
-				log.Tracef("sent flow event: %s", event.ID)
 			}
 		}
 	}
 }
 
-func (m *Manager) receiveACKs(ctx context.Context, client *client.GRPCClient) {
-	err := client.Receive(ctx, m.flowConfig.Interval, func(ack *proto.FlowEventAck) error {
+func (m *Manager) receiveACKs(ctx context.Context, client *client.GRPCClient, flowConfigInterval time.Duration) {
+	err := client.Receive(ctx, flowConfigInterval, func(ack *proto.FlowEventAck) error {
 		id, err := uuid.FromBytes(ack.EventId)
 		if err != nil {
 			log.Warnf("failed to convert ack event id to uuid: %v", err)
 			return nil
 		}
 		log.Tracef("received flow event ack: %s", id)
-		m.logger.DeleteEvents([]uuid.UUID{id})
+		m.eventsWithoutAcks.DeleteEvents([]uuid.UUID{id})
 		return nil
 	})
 
@@ -236,6 +251,51 @@ func (m *Manager) receiveACKs(ctx context.Context, client *client.GRPCClient) {
 	}
 }
 
+// We effectively never drop events (see MaxInterval), which makes eventsWithoutAcks unbounded.
+// We may want to limit the max size of the store, and start dropping oldest events when the threshold is reached.
+func (m *Manager) startRetries(ctx context.Context, flowConfigInterval time.Duration) {
+	timer := time.NewTimer(m.retryInterval)
+	retryBackoff := backoff.WithContext(&backoff.ExponentialBackOff{
+		InitialInterval:     1 * time.Second,
+		RandomizationFactor: 0.5,
+		Multiplier:          1.7,
+		MaxInterval:         flowConfigInterval / 2,
+		MaxElapsedTime:      3 * 30 * 24 * time.Hour, // 3 months
+		Stop:                backoff.Stop,
+		Clock:               backoff.SystemClock,
+	}, ctx)
+	defer timer.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-timer.C:
+			resetBackoff := true
+			for _, e := range m.eventsWithoutAcks.GetEvents() {
+				if e.Timestamp.Add(time.Second).After(time.Now()) {
+					// grace period on retries to avoid early retries
+					// do not retry if the event is less than 1 sec old
+					continue
+				}
+				if err := m.send(e); err != nil {
+					if nextBackoff := retryBackoff.NextBackOff(); nextBackoff != backoff.Stop {
+						timer = time.NewTimer(nextBackoff)
+						resetBackoff = false
+					} else {
+						resetBackoff = true // we exhausted retries, reset retry loop
+					}
+					break
+				}
+			}
+			if resetBackoff { // use regular retry interval in absence of network errors
+				retryBackoff.Reset()
+				timer = time.NewTimer(m.retryInterval)
+			}
+		}
+	}
+}
+
 func (m *Manager) send(event *nftypes.Event) error {
 	m.mux.Lock()
 	client := m.receiverClient
@@ -250,9 +310,11 @@ func (m *Manager) send(event *nftypes.Event) error {
 
 func toProtoEvent(publicKey []byte, event *nftypes.Event) *proto.FlowEvent {
 	protoEvent := &proto.FlowEvent{
-		EventId:   event.ID[:],
-		Timestamp: timestamppb.New(event.Timestamp),
-		PublicKey: publicKey,
+		EventId:     event.ID[:],
+		Timestamp:   timestamppb.New(event.Timestamp),
+		PublicKey:   publicKey,
+		WindowStart: timestamppb.New(event.WindowStart),
+		WindowEnd:   timestamppb.New(event.WindowEnd),
 		FlowFields: &proto.FlowFields{
 			FlowId:           event.FlowID[:],
 			RuleId:           event.RuleID,
@@ -267,6 +329,9 @@ func toProtoEvent(publicKey []byte, event *nftypes.Event) *proto.FlowEvent {
 			TxBytes:          event.TxBytes,
 			SourceResourceId: event.SourceResourceID,
 			DestResourceId:   event.DestResourceID,
+			NumOfStarts:      event.NumOfStarts,
+			NumOfEnds:        event.NumOfEnds,
+			NumOfDrops:       event.NumOfDrops,
 		},
 	}
 

client/internal/netflow/manager_integration_test.go

@@ -0,0 +1,291 @@
+package netflow
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"net"
+	"net/netip"
+	"slices"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/netbirdio/netbird/client/iface/wgaddr"
+	"github.com/netbirdio/netbird/client/internal/netflow/types"
+	"github.com/netbirdio/netbird/flow/proto"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"google.golang.org/grpc"
+)
+
+type testServer struct {
+	proto.UnimplementedFlowServiceServer
+	events         chan *proto.FlowEvent
+	acks           chan *proto.FlowEventAck
+	grpcSrv        *grpc.Server
+	addr           string
+	handlerDone    chan struct{} // signaled each time Events() exits
+	handlerStarted chan struct{} // signaled each time Events() begins
+}
+
+func newTestServer(t *testing.T) *testServer {
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+
+	s := &testServer{
+		events:         make(chan *proto.FlowEvent, 100),
+		acks:           make(chan *proto.FlowEventAck, 100),
+		grpcSrv:        grpc.NewServer(),
+		addr:           listener.Addr().String(),
+		handlerDone:    make(chan struct{}, 10),
+		handlerStarted: make(chan struct{}, 10),
+	}
+
+	proto.RegisterFlowServiceServer(s.grpcSrv, s)
+
+	go func() {
+		if err := s.grpcSrv.Serve(listener); err != nil && !errors.Is(err, grpc.ErrServerStopped) {
+			t.Logf("server error: %v", err)
+		}
+	}()
+
+	t.Cleanup(func() {
+		s.grpcSrv.Stop()
+	})
+
+	return s
+}
+
+func (s *testServer) Events(stream proto.FlowService_EventsServer) error {
+	defer func() {
+		select {
+		case s.handlerDone <- struct{}{}:
+		default:
+		}
+	}()
+
+	err := stream.Send(&proto.FlowEventAck{IsInitiator: true})
+	if err != nil {
+		return err
+	}
+
+	select {
+	case s.handlerStarted <- struct{}{}:
+	default:
+	}
+
+	ctx, cancel := context.WithCancel(stream.Context())
+	defer cancel()
+
+	go func() {
+		defer cancel()
+		for {
+			event, err := stream.Recv()
+			if err != nil {
+				return
+			}
+
+			if !event.IsInitiator {
+				select {
+				case s.events <- event:
+				case <-ctx.Done():
+					return
+				}
+			}
+		}
+	}()
+
+	for {
+		select {
+		case ack := <-s.acks:
+			if err := stream.Send(ack); err != nil {
+				return err
+			}
+		case <-ctx.Done():
+			return ctx.Err()
+		}
+	}
+}
+
+func TestSendEventReceiveAck(t *testing.T) {
+	_, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	t.Cleanup(cancel)
+
+	server := newTestServer(t)
+	manager := createManager(t, server.addr, 60*time.Second) // set high to prevent retries in this test
+	defer manager.Close()
+
+	assert.Eventually(t, func() bool {
+		select {
+		case <-server.handlerStarted:
+			return true
+		default:
+			return false
+		}
+	}, 3*time.Second, 100*time.Millisecond)
+
+	event1 := types.EventFields{
+		FlowID:    uuid.New(),
+		Type:      types.TypeStart,
+		Direction: types.Ingress,
+		DestIP:    ipAddr("172.16.1.2"),
+		DestPort:  2345,
+		Protocol:  6,
+	}
+	manager.logger.StoreEvent(event1)
+	event2 := types.EventFields{
+		FlowID:    uuid.New(),
+		Type:      types.TypeStart,
+		Direction: types.Ingress,
+		DestIP:    ipAddr("172.16.1.1"),
+		DestPort:  1234,
+		Protocol:  6,
+	}
+	manager.logger.StoreEvent(event2)
+
+	// verify the server received logged events
+	serverSideEvents := make([]*proto.FlowEvent, 0)
+	assert.Eventually(t, func() bool {
+		select {
+		case event := <-server.events:
+			serverSideEvents = append(serverSideEvents, event)
+			if len(serverSideEvents) == 2 {
+				return true
+			}
+		default:
+			if len(serverSideEvents) == 2 {
+				return true
+			}
+		}
+		return false
+	}, 5*time.Second, 100*time.Millisecond)
+
+	serverSideFlowIds := make([]uuid.UUID, 0, 2)
+	slices.Values(serverSideEvents)(func(e *proto.FlowEvent) bool {
+		id, err := uuid.FromBytes(e.FlowFields.FlowId)
+		assert.NoError(t, err)
+		serverSideFlowIds = append(serverSideFlowIds, id)
+		return true
+	})
+	assert.ElementsMatch(t, []uuid.UUID{event1.FlowID, event2.FlowID}, serverSideFlowIds)
+
+	// verify the manager tracks un-acked events
+	unackedEvents := manager.eventsWithoutAcks.GetEvents()
+	assert.Len(t, unackedEvents, 2)
+	flowIds := make([]uuid.UUID, 0)
+	slices.Values(unackedEvents)(func(e *types.Event) bool {
+		flowIds = append(flowIds, e.FlowID)
+		return true
+	})
+	assert.ElementsMatch(t, flowIds, []uuid.UUID{event1.FlowID, event2.FlowID})
+}
+
+// verify handling of retries:
+//   - unacked events are retried
+//   - when acks arrive, events are removed from the un-acked event tracker
+func TestRetryEvents(t *testing.T) {
+	_, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	t.Cleanup(cancel)
+
+	server := newTestServer(t)
+	manager := createManager(t, server.addr, time.Second) // set low to start retries sooner
+	defer manager.Close()
+
+	assert.Eventually(t, func() bool {
+		select {
+		case <-server.handlerStarted:
+			return true
+		default:
+			return false
+		}
+	}, 3*time.Second, 100*time.Millisecond)
+
+	event1 := types.EventFields{
+		FlowID:    uuid.New(),
+		Type:      types.TypeStart,
+		Direction: types.Ingress,
+		DestIP:    ipAddr("172.16.1.2"),
+		DestPort:  2345,
+		Protocol:  6,
+	}
+	manager.logger.StoreEvent(event1)
+	event2 := types.EventFields{
+		FlowID:    uuid.New(),
+		Type:      types.TypeStart,
+		Direction: types.Ingress,
+		DestIP:    ipAddr("172.16.1.1"),
+		DestPort:  1234,
+		Protocol:  6,
+	}
+	manager.logger.StoreEvent(event2)
+
+	// verify the server received retries of logged events
+	serverSideEvents := make([]*proto.FlowEvent, 0)
+	func() {
+		c := time.After(2500 * time.Millisecond)
+		for {
+			select {
+			case event := <-server.events:
+				serverSideEvents = append(serverSideEvents, event)
+			case <-c:
+				return
+			}
+		}
+	}()
+	assert.True(t, len(serverSideEvents) > 2) // must see retries
+
+	uniqueServerSideEvents := make(map[uuid.UUID]*proto.FlowEvent)
+	slices.Values(serverSideEvents)(func(e *proto.FlowEvent) bool {
+		id, err := uuid.FromBytes(e.FlowFields.FlowId)
+		assert.NoError(t, err)
+		uniqueServerSideEvents[id] = e
+		return true
+	})
+	assert.Contains(t, uniqueServerSideEvents, event1.FlowID)
+	assert.Contains(t, uniqueServerSideEvents, event2.FlowID)
+
+	// ack events
+	server.acks <- &proto.FlowEventAck{EventId: uniqueServerSideEvents[event1.FlowID].EventId}
+	server.acks <- &proto.FlowEventAck{EventId: uniqueServerSideEvents[event2.FlowID].EventId}
+
+	assert.EventuallyWithT(t, func(c *assert.CollectT) {
+		unackedEvents := manager.eventsWithoutAcks.GetEvents()
+		assert.Empty(c, unackedEvents)
+
+	}, 3*time.Second, 100*time.Millisecond)
+}
+
+func createManager(t *testing.T, serverAddr string, retryInterval time.Duration) *Manager {
+	t.Helper()
+
+	mockIFace := &mockIFaceMapper{
+		address: wgaddr.Address{
+			Network: netip.MustParsePrefix("192.168.1.1/32"),
+		},
+		isUserspaceBind: true,
+	}
+
+	publicKey := []byte("test-public-key")
+	manager := NewManager(mockIFace, publicKey, nil)
+	manager.retryInterval = retryInterval
+
+	initialConfig := &types.FlowConfig{
+		Enabled:        true,
+		URL:            fmt.Sprintf("http://%s", serverAddr),
+		TokenPayload:   "initial-payload",
+		TokenSignature: "initial-signature",
+		Interval:       500 * time.Millisecond,
+	}
+
+	err := manager.Update(initialConfig)
+	require.NoError(t, err)
+
+	return manager
+}
+
+func ipAddr(a string) netip.Addr {
+	addr, _ := netip.ParseAddr(a)
+	return addr
+}

client/internal/netflow/store/event_aggregation_test.go

@@ -0,0 +1,365 @@
+package store
+
+import (
+	"math/rand"
+	"net/netip"
+	"testing"
+	"time"
+
+	"github.com/google/uuid"
+	"github.com/netbirdio/netbird/client/internal/netflow/types"
+	"github.com/stretchr/testify/assert"
+)
+
+var random = rand.New(rand.NewSource(time.Now().UnixNano()))
+
+func TestFlowAggregation(t *testing.T) {
+	var protocols = []types.Protocol{types.ICMP, types.ICMPv6, types.TCP, types.UDP}
+	var tests = []struct {
+		description string
+		addresses   [][]netip.Addr
+		dstPort     uint16
+		eventTypes  []types.Type
+	}{
+		{
+			description: "start and stop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart, types.TypeEnd},
+		},
+		{
+			description: "start and drop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart, types.TypeDrop},
+		},
+		{
+			description: "start only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart},
+		},
+		{
+			description: "drop only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeDrop},
+		}}
+
+	for _, protocol := range protocols {
+		for _, tt := range tests {
+			t.Run(tt.description+" "+protocol.String(), func(t *testing.T) {
+				store := NewAggregatingMemoryStore()
+				store.WindowEnd = time.Now().Add(5 * time.Second)
+
+				allExpected := make([]*types.Event, 0)
+
+				for _, srcAndDst := range tt.addresses {
+					inEvents, expected := generateEvents(srcAndDst[0], srcAndDst[1], tt.dstPort, tt.eventTypes, protocol, types.Ingress, 0, store.WindowStart, store.WindowEnd)
+					for _, e := range inEvents {
+						store.StoreEvent(e)
+					}
+					allExpected = append(allExpected, expected)
+				}
+
+				events := store.GetAggregatedEvents()
+				assert.ElementsMatch(t, events, allExpected)
+			})
+		}
+	}
+}
+
+func TestIcmpEventAggregation(t *testing.T) {
+	var protocols = []types.Protocol{types.ICMP, types.ICMPv6}
+	var icmpTypes = []uint8{1, 2, 3}
+
+	var tests = []struct {
+		description string
+		addresses   [][]netip.Addr
+		eventTypes  []types.Type
+	}{
+		{
+			description: "start and stop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}},
+			eventTypes:  []types.Type{types.TypeStart, types.TypeEnd},
+		},
+		{
+			description: "start and drop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}},
+			eventTypes:  []types.Type{types.TypeStart, types.TypeDrop},
+		},
+		{
+			description: "start only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}},
+			eventTypes:  []types.Type{types.TypeStart},
+		},
+		{
+			description: "drop only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}},
+			eventTypes:  []types.Type{types.TypeDrop},
+		}}
+
+	for _, protocol := range protocols {
+		for _, tt := range tests {
+			t.Run(tt.description+" "+protocol.String(), func(t *testing.T) {
+				store := NewAggregatingMemoryStore()
+				store.WindowEnd = time.Now().Add(5 * time.Second)
+
+				allExpected := make([]*types.Event, 0)
+				for _, icmpType := range icmpTypes {
+					events, expected := generateEvents(tt.addresses[0][0], tt.addresses[0][1], 0, tt.eventTypes, protocol, types.Ingress, icmpType, store.WindowStart, store.WindowEnd)
+					for _, e := range events {
+						store.StoreEvent(e)
+					}
+					allExpected = append(allExpected, expected)
+				}
+				aggregatedEvents := store.GetAggregatedEvents()
+				assert.Len(t, aggregatedEvents, len(allExpected))
+				assert.ElementsMatch(t, aggregatedEvents, allExpected)
+			})
+		}
+	}
+}
+
+func TestFlowAggregationOfUnknownProtocols(t *testing.T) {
+	var tests = []struct {
+		description string
+		addresses   [][]netip.Addr
+		dstPort     uint16
+		eventTypes  []types.Type
+	}{
+		{
+			description: "start and stop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart, types.TypeEnd},
+		},
+		{
+			description: "start and drop",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart, types.TypeDrop},
+		},
+		{
+			description: "start only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeStart},
+		},
+		{
+			description: "drop only",
+			addresses:   [][]netip.Addr{{netip.MustParseAddr("1.1.1.1"), netip.MustParseAddr("2.2.2.2")}, {netip.MustParseAddr("3.3.3.3"), netip.MustParseAddr("2.2.2.2")}},
+			dstPort:     uint16(random.Uint32() >> 16),
+			eventTypes:  []types.Type{types.TypeDrop},
+		}}
+
+	for _, tt := range tests {
+		t.Run(tt.description+" "+types.ProtocolUnknown.String(), func(t *testing.T) {
+			store := NewAggregatingMemoryStore()
+			store.WindowEnd = time.Now().Add(5 * time.Second)
+
+			allExpected := make([]*types.Event, 0)
+
+			for _, srcAndDst := range tt.addresses {
+				inEvents, expected := generateEventsForUnknownProtocol(srcAndDst[0], srcAndDst[1], tt.dstPort, tt.eventTypes, types.ProtocolUnknown, types.Ingress, store.WindowStart, store.WindowEnd)
+				for _, e := range inEvents {
+					store.StoreEvent(e)
+				}
+				allExpected = append(allExpected, expected...)
+			}
+
+			events := store.GetAggregatedEvents()
+			assert.ElementsMatch(t, events, allExpected)
+		})
+	}
+}
+
+func TestResetAggregationWindow(t *testing.T) {
+	now := time.Now()
+	nowFunc := func() time.Time { return now }
+	store := NewAggregatingMemoryStoreWithTimeFunc(nowFunc)
+	store.StoreEvent(&types.Event{
+		ID:        uuid.New(),
+		Timestamp: time.Now(),
+		EventFields: types.EventFields{
+			FlowID:           uuid.New(),
+			Type:             types.TypeStart,
+			Protocol:         types.TCP,
+			RuleID:           []byte("rule-id-1"),
+			Direction:        types.Ingress,
+			SourceIP:         netip.MustParseAddr("1.1.1.1"),
+			SourcePort:       1234,
+			DestIP:           netip.MustParseAddr("2.2.2.2"),
+			DestPort:         5678,
+			SourceResourceID: []byte("source-resource-id"),
+			DestResourceID:   []byte("dest-resource-id"),
+			RxPackets:        random.Uint64(),
+			TxPackets:        random.Uint64(),
+			RxBytes:          random.Uint64(),
+			TxBytes:          random.Uint64(),
+		},
+	})
+
+	now = now.Add(1 * time.Second)
+	reset := store.ResetAggregationWindow()
+	previousEvents, ok := reset.(*AggregatingMemory)
+	assert.True(t, ok)
+	assert.NotEqual(t, previousEvents.WindowStart, store.WindowStart)
+	assert.Equal(t, previousEvents.WindowEnd, store.WindowStart)
+	assert.NotEmpty(t, previousEvents.events)
+	assert.Empty(t, store.events)
+}
+
+func generateEvents(srcIp, dstIp netip.Addr, dstPort uint16, eventTypes []types.Type, protocol types.Protocol,
+	direction types.Direction, icmpType uint8, windowStart, windowEnd time.Time) ([]*types.Event, *types.Event) {
+	var rxPackets, txPackets, rxBytes, txBytes uint64
+	inEvents := make([]*types.Event, 0)
+	ts := time.Now()
+	flowId := uuid.New()
+	srcPort := uint16(random.Uint32() >> 16)
+
+	for idx, eventType := range eventTypes {
+		e := &types.Event{
+			ID:        uuid.New(),
+			Timestamp: ts.Add(time.Duration(idx) * time.Second),
+			EventFields: types.EventFields{
+				FlowID:           flowId,
+				Type:             eventType,
+				Protocol:         protocol,
+				RuleID:           []byte("rule-id-1"),
+				Direction:        direction,
+				SourceIP:         srcIp,
+				SourcePort:       srcPort,
+				DestIP:           dstIp,
+				DestPort:         dstPort,
+				SourceResourceID: []byte("source-resource-id"),
+				DestResourceID:   []byte("dest-resource-id"),
+				RxPackets:        random.Uint64(),
+				TxPackets:        random.Uint64(),
+				RxBytes:          random.Uint64(),
+				TxBytes:          random.Uint64(),
+			}}
+		rxBytes += e.RxBytes
+		txBytes += e.TxBytes
+		rxPackets += e.RxPackets
+		txPackets += e.TxPackets
+		inEvents = append(inEvents, e)
+		if protocol == types.ICMP || protocol == types.ICMPv6 {
+			e.ICMPType = icmpType
+		}
+	}
+
+	var start, end, drop uint64
+	for _, eventType := range eventTypes {
+		switch eventType {
+		case types.TypeStart:
+			start += 1
+		case types.TypeDrop:
+			drop += 1
+		case types.TypeEnd:
+			end += 1
+		}
+	}
+	aggregatedEvent := &types.Event{
+		ID:          inEvents[0].ID,
+		Timestamp:   inEvents[0].Timestamp,
+		WindowStart: windowStart,
+		WindowEnd:   windowEnd,
+		EventFields: types.EventFields{
+			FlowID:           flowId,
+			Type:             types.TypeUnknown,
+			Protocol:         inEvents[0].Protocol,
+			RuleID:           []byte("rule-id-1"),
+			Direction:        inEvents[0].Direction,
+			SourceIP:         srcIp,
+			SourcePort:       srcPort,
+			DestIP:           dstIp,
+			DestPort:         dstPort,
+			SourceResourceID: []byte("source-resource-id"),
+			DestResourceID:   []byte("dest-resource-id"),
+			RxPackets:        rxPackets,
+			TxPackets:        txPackets,
+			RxBytes:          rxBytes,
+			TxBytes:          txBytes,
+			NumOfStarts:      start,
+			NumOfEnds:        end,
+			NumOfDrops:       drop,
+		}}
+	if protocol == types.ICMP || protocol == types.ICMPv6 {
+		aggregatedEvent.ICMPType = icmpType
+	}
+
+	return inEvents, aggregatedEvent
+}
+
+func generateEventsForUnknownProtocol(srcIp, dstIp netip.Addr, dstPort uint16, eventTypes []types.Type, protocol types.Protocol,
+	direction types.Direction, windowStart, windowEnd time.Time) ([]*types.Event, []*types.Event) {
+	inEvents := make([]*types.Event, 0)
+	expectedEvents := make([]*types.Event, 0)
+
+	ts := time.Now()
+	flowId := uuid.New()
+	srcPort := uint16(random.Uint32() >> 16)
+
+	for idx, eventType := range eventTypes {
+		e := &types.Event{
+			ID:        uuid.New(),
+			Timestamp: ts.Add(time.Duration(idx) * time.Second),
+			EventFields: types.EventFields{
+				FlowID:           flowId,
+				Type:             eventType,
+				Protocol:         protocol,
+				RuleID:           []byte("rule-id-1"),
+				Direction:        direction,
+				SourceIP:         srcIp,
+				SourcePort:       srcPort,
+				DestIP:           dstIp,
+				DestPort:         dstPort,
+				SourceResourceID: []byte("source-resource-id"),
+				DestResourceID:   []byte("dest-resource-id"),
+				RxPackets:        random.Uint64(),
+				TxPackets:        random.Uint64(),
+				RxBytes:          random.Uint64(),
+				TxBytes:          random.Uint64(),
+			}}
+		inEvents = append(inEvents, e)
+
+		var start, end, drop uint64
+		switch eventType {
+		case types.TypeStart:
+			start = 1
+		case types.TypeDrop:
+			drop = 1
+		case types.TypeEnd:
+			end = 1
+		}
+
+		expectedEvents = append(expectedEvents, &types.Event{
+			ID:          e.ID,
+			Timestamp:   e.Timestamp,
+			WindowStart: windowStart,
+			WindowEnd:   windowEnd,
+			EventFields: types.EventFields{
+				FlowID:           flowId,
+				Type:             types.TypeUnknown,
+				Protocol:         e.Protocol,
+				RuleID:           []byte("rule-id-1"),
+				Direction:        e.Direction,
+				SourceIP:         srcIp,
+				SourcePort:       srcPort,
+				DestIP:           dstIp,
+				DestPort:         dstPort,
+				SourceResourceID: []byte("source-resource-id"),
+				DestResourceID:   []byte("dest-resource-id"),
+				RxPackets:        e.RxPackets,
+				TxPackets:        e.TxPackets,
+				RxBytes:          e.RxBytes,
+				TxBytes:          e.TxBytes,
+				NumOfStarts:      start,
+				NumOfEnds:        end,
+				NumOfDrops:       drop,
+			}})
+	}
+
+	return inEvents, expectedEvents
+}

client/internal/netflow/store/memory.go

@@ -1,10 +1,15 @@
 package store
 
 import (
+	"maps"
+	"math/rand"
+	v2 "math/rand/v2"
+	"net/netip"
+	"slices"
 	"sync"
+	"time"
 
 	"github.com/google/uuid"
-
 	"github.com/netbirdio/netbird/client/internal/netflow/types"
 )
 
@@ -19,6 +24,14 @@ type Memory struct {
 	events map[uuid.UUID]*types.Event
 }
 
+type AggregatingMemory struct {
+	Memory
+	WindowStart time.Time
+	WindowEnd   time.Time
+	rnd         *v2.PCG
+	nowFunc     func() time.Time
+}
+
 func (m *Memory) StoreEvent(event *types.Event) {
 	m.mux.Lock()
 	defer m.mux.Unlock()
@@ -48,3 +61,104 @@ func (m *Memory) DeleteEvents(ids []uuid.UUID) {
 		delete(m.events, id)
 	}
 }
+
+func NewAggregatingMemoryStore() *AggregatingMemory {
+	return NewAggregatingMemoryStoreWithTimeFunc(defaultNowFunc)
+}
+
+// used in tests when deterministic (less random) time intervals are required
+func NewAggregatingMemoryStoreWithTimeFunc(nowFunc func() time.Time) *AggregatingMemory {
+	return &AggregatingMemory{WindowStart: nowFunc(), Memory: Memory{events: make(map[uuid.UUID]*types.Event)}, nowFunc: nowFunc, rnd: v2.NewPCG(rand.Uint64(), rand.Uint64())}
+}
+
+func (am *AggregatingMemory) ResetAggregationWindow() types.FlowEventAggregator {
+	am.mux.Lock()
+	defer am.mux.Unlock()
+
+	now := am.nowFunc()
+	toret := AggregatingMemory{WindowStart: am.WindowStart, WindowEnd: now, Memory: Memory{events: am.events}, rnd: v2.NewPCG(rand.Uint64(), rand.Uint64())}
+
+	am.events = make(map[uuid.UUID]*types.Event)
+	am.WindowStart = now
+
+	return &toret
+}
+
+type aggregationKey struct {
+	srcAddr   netip.Addr
+	destAddr  netip.Addr
+	destPort  uint16
+	direction int
+	protocol  uint8
+	icmpType  uint8
+	unique    uint64 // used to prevent aggregation on non icmp/udp/tcp events
+}
+
+func (am *AggregatingMemory) GetAggregatedEvents() []*types.Event {
+	am.mux.Lock()
+	defer am.mux.Unlock()
+
+	aggregated := make(map[aggregationKey]*types.Event)
+	for _, v := range am.events {
+		lookupKey := aggregationKey{srcAddr: v.SourceIP, destAddr: v.DestIP, destPort: v.DestPort, direction: int(v.Direction), protocol: uint8(v.Protocol), icmpType: v.ICMPType}
+		if _, ok := aggregated[lookupKey]; !ok {
+			event := v.Clone()
+
+			switch event.Type {
+			case types.TypeStart:
+				event.NumOfStarts += 1
+			case types.TypeDrop:
+				event.NumOfDrops += 1
+			case types.TypeEnd:
+				event.NumOfEnds += 1
+			}
+			event.Type = types.TypeUnknown
+
+			// Please note that ICMPCode field isn't propagated by the manager (see flow/proto/flow.pb.go, FlowFields struct)
+			// so the field value in an icmp event in the "aggregated" doesn't matter
+
+			event.WindowStart = am.WindowStart
+			event.WindowEnd = am.WindowEnd
+
+			if event.Protocol != types.ICMP && event.Protocol != types.ICMPv6 && event.Protocol != types.UDP && event.Protocol != types.TCP {
+				lookupKey.unique = am.rnd.Uint64() // to make the lookup key unique so we don't aggregate on it
+			}
+
+			aggregated[lookupKey] = event
+			continue
+		}
+
+		aggregatedEvent := aggregated[lookupKey]
+		if aggregatedEvent.Protocol != types.ICMP && aggregatedEvent.Protocol != types.ICMPv6 && aggregatedEvent.Protocol != types.UDP && aggregatedEvent.Protocol != types.TCP {
+			continue // we don't aggregate this type of events; shouldn't ever get here
+		}
+
+		// track the number of connections, duration?, open and close events?
+		aggregatedEvent.RxBytes += v.RxBytes
+		aggregatedEvent.RxPackets += v.RxPackets
+		aggregatedEvent.TxBytes += v.TxBytes
+		aggregatedEvent.TxPackets += v.TxPackets
+		switch v.Type {
+		case types.TypeStart:
+			aggregatedEvent.NumOfStarts += 1
+		case types.TypeDrop:
+			aggregatedEvent.NumOfDrops += 1
+		case types.TypeEnd:
+			aggregatedEvent.NumOfEnds += 1
+		}
+		if aggregatedEvent.Timestamp.Compare(v.Timestamp) > 0 {
+			aggregatedEvent.Timestamp = v.Timestamp
+			aggregatedEvent.ID = v.ID
+			aggregatedEvent.SourcePort = v.SourcePort
+		}
+		if len(aggregatedEvent.RuleID) == 0 && len(v.RuleID) != 0 {
+			aggregatedEvent.RuleID = slices.Clone(v.RuleID)
+		}
+	}
+
+	return slices.Collect(maps.Values(aggregated)) // could return an iterator instead here
+}
+
+func defaultNowFunc() time.Time {
+	return time.Now()
+}

client/internal/netflow/types/types.go

@@ -2,6 +2,7 @@ package types
 
 import (
 	"net/netip"
+	"slices"
 	"strconv"
 	"time"
 
@@ -69,8 +70,10 @@ const (
 )
 
 type Event struct {
-	ID        uuid.UUID
-	Timestamp time.Time
+	ID          uuid.UUID
+	Timestamp   time.Time
+	WindowStart time.Time
+	WindowEnd   time.Time
 	EventFields
 }
 
@@ -92,6 +95,17 @@ type EventFields struct {
 	TxPackets        uint64
 	RxBytes          uint64
 	TxBytes          uint64
+	NumOfStarts      uint64
+	NumOfEnds        uint64
+	NumOfDrops       uint64
+}
+
+func (e *Event) Clone() *Event {
+	toret := *e
+	toret.RuleID = slices.Clone(e.RuleID)
+	toret.SourceResourceID = slices.Clone(e.SourceResourceID)
+	toret.DestResourceID = slices.Clone(e.DestResourceID)
+	return &toret
 }
 
 type FlowConfig struct {
@@ -114,13 +128,15 @@ type FlowManager interface {
 	GetLogger() FlowLogger
 }
 
+type FlowEventAggregator interface {
+	ResetAggregationWindow() FlowEventAggregator
+	GetAggregatedEvents() []*Event
+}
+
 type FlowLogger interface {
+	ResetAggregationWindow() FlowEventAggregator
 	// StoreEvent stores a flow event
 	StoreEvent(flowEvent EventFields)
-	// GetEvents returns all stored events
-	GetEvents() []*Event
-	// DeleteEvents deletes events from the store
-	DeleteEvents([]uuid.UUID)
 	// Close closes the logger
 	Close()
 	// Enable enables the flow logger receiver
@@ -140,6 +156,11 @@ type Store interface {
 	Close()
 }
 
+type AggregatingStore interface {
+	FlowEventAggregator
+	Store
+}
+
 // ConnTracker defines the interface for connection tracking functionality
 type ConnTracker interface {
 	// Start begins tracking connections by listening for conntrack events.

client/internal/peer/conn_status.go

@@ -1,18 +1,5 @@
 package peer
 
-import (
-	log "github.com/sirupsen/logrus"
-)
-
-const (
-	// StatusIdle indicate the peer is in disconnected state
-	StatusIdle ConnStatus = iota
-	// StatusConnecting indicate the peer is in connecting state
-	StatusConnecting
-	// StatusConnected indicate the peer is in connected state
-	StatusConnected
-)
-
 // connStatusInputs is the primitive-valued snapshot of the state that drives the
 // tri-state connection classification. Extracted so the decision logic can be unit-tested
 // without constructing full Worker/Handshaker objects.
@@ -21,24 +8,7 @@ type connStatusInputs struct {
 	peerUsesRelay       bool // remote peer advertises relay support AND local has relay
 	relayConnected      bool // statusRelay reports Connected (independent of whether peer uses relay)
 	remoteSupportsICE   bool // remote peer sent ICE credentials
-	iceWorkerCreated    bool // local WorkerICE exists (false in force-relay mode)
+	iceWorkerCreated    bool // local ICE worker exists (false in force-relay mode)
 	iceStatusConnecting bool // statusICE is anything other than Disconnected
 	iceInProgress       bool // a negotiation is currently in flight
 }
-
-// ConnStatus describe the status of a peer's connection
-type ConnStatus int32
-
-func (s ConnStatus) String() string {
-	switch s {
-	case StatusConnecting:
-		return "Connecting"
-	case StatusConnected:
-		return "Connected"
-	case StatusIdle:
-		return "Idle"
-	default:
-		log.Errorf("unknown status: %d", s)
-		return "INVALID_PEER_CONNECTION_STATUS"
-	}
-}

client/internal/peer/conn_test.go

@@ -3,27 +3,33 @@ package peer
 import (
 	"context"
 	"fmt"
+	"net/netip"
 	"os"
 	"testing"
 	"time"
 
+	log "github.com/sirupsen/logrus"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 
 	"github.com/netbirdio/netbird/client/iface"
-	"github.com/netbirdio/netbird/client/internal/peer/dispatcher"
 	"github.com/netbirdio/netbird/client/internal/peer/guard"
 	"github.com/netbirdio/netbird/client/internal/peer/ice"
+	"github.com/netbirdio/netbird/client/internal/peer/metricsstages"
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
+	"github.com/netbirdio/netbird/client/internal/peer/status"
 	"github.com/netbirdio/netbird/client/internal/stdnet"
 	"github.com/netbirdio/netbird/util"
 )
 
-var testDispatcher = dispatcher.NewConnectionDispatcher()
-
 var connConf = ConnConfig{
 	Key:         "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
 	LocalKey:    "RRHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
 	Timeout:     time.Second,
 	LocalWgPort: 51820,
+	WgConfig: WgConfig{
+		AllowedIps: []netip.Prefix{netip.MustParsePrefix("100.64.0.1/32")},
+	},
 	ICEConfig: ice.Config{
 		InterfaceBlackList: nil,
 	},
@@ -51,92 +57,37 @@ func TestConn_GetKey(t *testing.T) {
 	swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
 
 	sd := ServiceDependencies{
-		SrWatcher:          swWatcher,
-		PeerConnDispatcher: testDispatcher,
+		SrWatcher: swWatcher,
 	}
 	conn, err := NewConn(connConf, sd)
-	if err != nil {
-		return
-	}
+	require.NoError(t, err)
 
 	got := conn.GetKey()
 
 	assert.Equal(t, got, connConf.Key, "they should be equal")
 }
 
-func TestConn_OnRemoteOffer(t *testing.T) {
+// TestConn_DiscardMessagesWhenNotOpened: signal messages posted to a not yet
+// opened connection must be discarded without blocking or panicking.
+func TestConn_DiscardMessagesWhenNotOpened(t *testing.T) {
 	swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
 	sd := ServiceDependencies{
-		StatusRecorder:     NewRecorder("https://mgm"),
-		SrWatcher:          swWatcher,
-		PeerConnDispatcher: testDispatcher,
+		StatusRecorder: status.NewRecorder("https://mgm"),
+		SrWatcher:      swWatcher,
 	}
 	conn, err := NewConn(connConf, sd)
-	if err != nil {
-		return
-	}
+	require.NoError(t, err)
 
-	onNewOfferChan := make(chan struct{})
-
-	conn.handshaker.AddRelayListener(func(remoteOfferAnswer *OfferAnswer) {
-		onNewOfferChan <- struct{}{}
-	})
-
-	conn.OnRemoteOffer(OfferAnswer{
-		IceCredentials: IceCredentials{
+	offerAnswer := signaling.OfferAnswer{
+		IceCredentials: signaling.IceCredentials{
 			UFrag: "test",
 			Pwd:   "test",
 		},
-		WgListenPort: 0,
-		Version:      "",
-	})
-
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-	defer cancel()
-
-	select {
-	case <-onNewOfferChan:
-		// success
-	case <-ctx.Done():
-		t.Error("expected to receive a new offer notification, but timed out")
-	}
-}
-
-func TestConn_OnRemoteAnswer(t *testing.T) {
-	swWatcher := guard.NewSRWatcher(nil, nil, nil, connConf.ICEConfig)
-	sd := ServiceDependencies{
-		StatusRecorder:     NewRecorder("https://mgm"),
-		SrWatcher:          swWatcher,
-		PeerConnDispatcher: testDispatcher,
-	}
-	conn, err := NewConn(connConf, sd)
-	if err != nil {
-		return
-	}
-
-	onNewOfferChan := make(chan struct{})
-
-	conn.handshaker.AddRelayListener(func(remoteOfferAnswer *OfferAnswer) {
-		onNewOfferChan <- struct{}{}
-	})
-
-	conn.OnRemoteAnswer(OfferAnswer{
-		IceCredentials: IceCredentials{
-			UFrag: "test",
-			Pwd:   "test",
-		},
-		WgListenPort: 0,
-		Version:      "",
-	})
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-	defer cancel()
-
-	select {
-	case <-onNewOfferChan:
-		// success
-	case <-ctx.Done():
-		t.Error("expected to receive a new offer notification, but timed out")
 	}
+	conn.OnRemoteOffer(offerAnswer)
+	conn.OnRemoteAnswer(offerAnswer)
+	conn.OnRemoteCandidate(nil, nil)
+	conn.Close(false)
 }
 
 func TestConn_presharedKey(t *testing.T) {
@@ -304,3 +255,84 @@ func TestConn_presharedKey_RosenpassManaged(t *testing.T) {
 		t.Fatalf("expected non-nil presharedKey before Rosenpass manages PSK")
 	}
 }
+
+func newWGTimeoutTestConn(rosenpassEnabled bool, disconnected *[]string) *Conn {
+	cfg := ConnConfig{
+		Key:      "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
+		LocalKey: "RRHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU=",
+		WgConfig: WgConfig{RemoteKey: "LLHf3Ma6z6mdLbriAJbqhX7+nM/B71lgw2+91q3LfhU="},
+	}
+	if rosenpassEnabled {
+		cfg.RosenpassConfig = RosenpassConfig{PubKey: []byte("dummykey")}
+	}
+
+	conn := &Conn{
+		ctx:           context.Background(),
+		config:        cfg,
+		Log:           log.WithField("peer", cfg.Key),
+		metricsStages: &metricsstages.MetricsStages{},
+	}
+	conn.SetOnDisconnected(func(remotePeer string) {
+		*disconnected = append(*disconnected, remotePeer)
+	})
+	return conn
+}
+
+// TestConn_onWGDisconnected_EscalatesToRosenpassReset: repeated handshake
+// timeouts with rosenpass enabled mean the preshared keys have desynced. The
+// renewal exchange runs over the dead tunnel and cannot resync them, so after
+// wgTimeoutEscalationThreshold consecutive timeouts the conn must report the
+// peer disconnected, dropping its rosenpass state so the next configuration
+// programs the rendezvous key.
+func TestConn_onWGDisconnected_EscalatesToRosenpassReset(t *testing.T) {
+	var disconnected []string
+	conn := newWGTimeoutTestConn(true, &disconnected)
+
+	for i := 0; i < wgTimeoutEscalationThreshold-1; i++ {
+		conn.handleWGTimeout()
+	}
+	assert.Empty(t, disconnected, "escalation must not fire below the threshold")
+
+	conn.handleWGTimeout()
+	assert.Equal(t, []string{conn.config.WgConfig.RemoteKey}, disconnected,
+		"reaching the threshold must report the peer disconnected once")
+
+	for i := 0; i < wgTimeoutEscalationThreshold-1; i++ {
+		conn.handleWGTimeout()
+	}
+	assert.Len(t, disconnected, 1, "escalation must restart counting after firing")
+
+	conn.handleWGTimeout()
+	assert.Len(t, disconnected, 2, "continued timeouts must escalate again")
+}
+
+// TestConn_onWGDisconnected_CheckSuccessResetsEscalation: a successful
+// handshake between timeouts means the tunnel recovered; the counter must
+// start over.
+func TestConn_onWGDisconnected_CheckSuccessResetsEscalation(t *testing.T) {
+	var disconnected []string
+	conn := newWGTimeoutTestConn(true, &disconnected)
+
+	for i := 0; i < wgTimeoutEscalationThreshold-1; i++ {
+		conn.handleWGTimeout()
+	}
+	conn.handleWGCheckSuccess()
+
+	for i := 0; i < wgTimeoutEscalationThreshold-1; i++ {
+		conn.handleWGTimeout()
+	}
+	assert.Empty(t, disconnected, "handshake success must reset the timeout count")
+}
+
+// TestConn_onWGDisconnected_NoEscalationWithoutRosenpass: without rosenpass
+// there is no per-peer key state to reset; repeated timeouts must not report
+// disconnects.
+func TestConn_onWGDisconnected_NoEscalationWithoutRosenpass(t *testing.T) {
+	var disconnected []string
+	conn := newWGTimeoutTestConn(false, &disconnected)
+
+	for i := 0; i < wgTimeoutEscalationThreshold*3; i++ {
+		conn.handleWGTimeout()
+	}
+	assert.Empty(t, disconnected, "escalation must be limited to rosenpass connections")
+}

client/internal/peer/dispatcher/dispatcher.go

@@ -1,52 +0,0 @@
-package dispatcher
-
-import (
-	"sync"
-
-	"github.com/netbirdio/netbird/client/internal/peer/id"
-)
-
-type ConnectionListener struct {
-	OnConnected    func(peerID id.ConnID)
-	OnDisconnected func(peerID id.ConnID)
-}
-
-type ConnectionDispatcher struct {
-	listeners map[*ConnectionListener]struct{}
-	mu        sync.Mutex
-}
-
-func NewConnectionDispatcher() *ConnectionDispatcher {
-	return &ConnectionDispatcher{
-		listeners: make(map[*ConnectionListener]struct{}),
-	}
-}
-
-func (e *ConnectionDispatcher) AddListener(listener *ConnectionListener) {
-	e.mu.Lock()
-	defer e.mu.Unlock()
-	e.listeners[listener] = struct{}{}
-}
-
-func (e *ConnectionDispatcher) RemoveListener(listener *ConnectionListener) {
-	e.mu.Lock()
-	defer e.mu.Unlock()
-
-	delete(e.listeners, listener)
-}
-
-func (e *ConnectionDispatcher) NotifyConnected(peerConnID id.ConnID) {
-	e.mu.Lock()
-	defer e.mu.Unlock()
-	for listener := range e.listeners {
-		listener.OnConnected(peerConnID)
-	}
-}
-
-func (e *ConnectionDispatcher) NotifyDisconnected(peerConnID id.ConnID) {
-	e.mu.Lock()
-	defer e.mu.Unlock()
-	for listener := range e.listeners {
-		listener.OnDisconnected(peerConnID)
-	}
-}

client/internal/peer/event.go

@@ -0,0 +1,69 @@
+package peer
+
+import (
+	"time"
+
+	"github.com/pion/ice/v4"
+
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
+	"github.com/netbirdio/netbird/client/internal/peer/worker"
+	"github.com/netbirdio/netbird/route"
+)
+
+// event is a message processed by the Conn event loop. All mutable Conn state
+// is owned by that loop; producers deliver events through the mailbox and
+// never mutate Conn state directly.
+type event any
+
+// evClose asks the event loop to tear down the connection. done is closed
+// once the teardown finished.
+type evClose struct {
+	signalToRemote bool
+	done           chan struct{}
+}
+
+type evRemoteOffer struct {
+	offer signaling.OfferAnswer
+}
+
+type evRemoteAnswer struct {
+	answer signaling.OfferAnswer
+}
+
+type evRemoteCandidate struct {
+	candidate ice.Candidate
+	haRoutes  route.HAMap
+}
+
+type evICEReady struct {
+	priority worker.ConnPriority
+	info     worker.ICEConnInfo
+}
+
+type evICEDown struct {
+	sessionChanged bool
+}
+
+type evRelayReady struct {
+	info worker.RelayConnInfo
+}
+
+type evRelayDown struct{}
+
+// evRelayDialDone reports that the relay dial helper goroutine finished,
+// successfully or not, so the loop may dispatch a pending offer.
+type evRelayDialDone struct{}
+
+type evWGTimeout struct{}
+
+// evWGHandshake reports the first WireGuard handshake of the current watcher run.
+type evWGHandshake struct {
+	when time.Time
+}
+
+// evWGCheckOK reports a watcher check that observed a fresh handshake,
+// including handshakes of connections that were already up.
+type evWGCheckOK struct{}
+
+// evGuardTick asks the loop to send a new offer to restore connectivity.
+type evGuardTick struct{}

client/internal/peer/guard/guard.go

@@ -85,7 +85,11 @@ func (g *Guard) reconnectLoopWithRetry(ctx context.Context, callback func()) {
 	defer g.srWatcher.RemoveListener(srReconnectedChan)
 
 	ticker := g.initialTicker(ctx)
-	defer ticker.Stop()
+	defer func() {
+		// If backoff.Ticker.send is blocked, context.Done will not close the Ticker goroutine.
+		// We have to explicitly call Stop, even if we use backoff.WithContext.
+		ticker.Stop()
+	}()
 
 	tickerChannel := ticker.C
 

client/internal/peer/guard/guard_leak_test.go

@@ -0,0 +1,92 @@
+package guard
+
+import (
+	"context"
+	"runtime"
+	"strings"
+	"sync"
+	"testing"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/peer/ice"
+)
+
+func newTestGuard(status connStatusFunc) *Guard {
+	srw := NewSRWatcher(nil, nil, nil, ice.Config{})
+	return NewGuard(log.WithField("test", "guard"), status, 50*time.Millisecond, srw)
+}
+
+// countBackoffTickerGoroutines returns how many goroutines are currently sitting
+// in backoff/v4.(*Ticker).run (a ticker goroutine that has not exited).
+func countBackoffTickerGoroutines() int {
+	buf := make([]byte, 1<<25) // 32MB
+	n := runtime.Stack(buf, true)
+	return strings.Count(string(buf[:n]), "backoff/v4.(*Ticker).run")
+}
+
+// TestGuard_ReconnectTicker_NoGoroutineLeakOnShutdown reproduces a observed
+// leak: after a shutdown burst, ticker run/send goroutines stay parked
+// forever even though every reconnect loop has exited.
+func TestGuard_ReconnectTicker_NoGoroutineLeakOnShutdown(t *testing.T) {
+	before := countBackoffTickerGoroutines()
+
+	const peers = 6000
+	cancels := make([]context.CancelFunc, 0, peers)
+	var wg sync.WaitGroup
+
+	// A status check slower than the tick cadence. This models the real
+	// isConnectedOnAllWay/callback doing work: while the loop is busy in the
+	// handler, the ticker fires the next tick and parks in send(), because
+	// send() never selects on ctx.
+	slowStatus := func() ConnStatus {
+		time.Sleep(70 * time.Millisecond)
+		return ConnStatusConnected
+	}
+
+	for range peers {
+		g := newTestGuard(slowStatus)
+		ctx, cancel := context.WithCancel(context.Background())
+		cancels = append(cancels, cancel)
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			g.Start(ctx, func() {})
+		}()
+		// Force the live ticker to be a newReconnectTicker.
+		g.SetRelayedConnDisconnected()
+	}
+
+	// Let the replacement tickers get past their 800ms initial interval, so
+	// many are parked in send() waiting on the (slow) consumer when we tear
+	// everything down.
+	time.Sleep(1500 * time.Millisecond)
+
+	// Shutdown burst: cancel every peer at once, like engine teardown.
+	for _, c := range cancels {
+		c()
+	}
+
+	// Every reconnect loop must return
+	waitCh := make(chan struct{})
+	go func() { wg.Wait(); close(waitCh) }()
+	select {
+	case <-waitCh:
+	case <-time.After(30 * time.Second):
+		t.Fatal("not all reconnect loops returned after ctx cancel")
+	}
+
+	// Give any correctly-stopped ticker goroutines time to unwind.
+	for range 50 {
+		runtime.Gosched()
+		time.Sleep(10 * time.Millisecond)
+	}
+
+	leaked := countBackoffTickerGoroutines() - before
+	t.Logf("backoff Ticker.run goroutines still parked after teardown of %d peers: %d", peers, leaked)
+	if leaked > 0 {
+		t.Errorf("LEAK: %d backoff ticker goroutines parked after all reconnect loops exited "+
+			"(defer ticker.Stop() stops the initial ticker, not the live replacement)", leaked)
+	}
+}

client/internal/peer/guard/ice_monitor.go

@@ -21,8 +21,6 @@ const (
 )
 
 type ICEMonitor struct {
-	ReconnectCh chan struct{}
-
 	iFaceDiscover stdnet.ExternalIFaceDiscover
 	iceConfig     icemaker.Config
 	tickerPeriod  time.Duration
@@ -34,7 +32,6 @@ type ICEMonitor struct {
 func NewICEMonitor(iFaceDiscover stdnet.ExternalIFaceDiscover, config icemaker.Config, period time.Duration) *ICEMonitor {
 	log.Debugf("prepare ICE monitor with period: %s", period)
 	cm := &ICEMonitor{
-		ReconnectCh:   make(chan struct{}, 1),
 		iFaceDiscover: iFaceDiscover,
 		iceConfig:     config,
 		tickerPeriod:  period,

client/internal/peer/handshaker.go

@@ -1,246 +0,0 @@
-package peer
-
-import (
-	"context"
-	"errors"
-	"net/netip"
-	"sync"
-	"sync/atomic"
-
-	log "github.com/sirupsen/logrus"
-
-	"github.com/netbirdio/netbird/version"
-)
-
-var (
-	ErrSignalIsNotReady = errors.New("signal is not ready")
-)
-
-// IceCredentials ICE protocol credentials struct
-type IceCredentials struct {
-	UFrag string
-	Pwd   string
-}
-
-// OfferAnswer represents a session establishment offer or answer
-type OfferAnswer struct {
-	IceCredentials IceCredentials
-	// WgListenPort is a remote WireGuard listen port.
-	// This field is used when establishing a direct WireGuard connection without any proxy.
-	// We can set the remote peer's endpoint with this port.
-	WgListenPort int
-
-	// Version of NetBird Agent
-	Version string
-	// RosenpassPubKey is the Rosenpass public key of the remote peer when receiving this message
-	// This value is the local Rosenpass server public key when sending the message
-	RosenpassPubKey []byte
-	// RosenpassAddr is the Rosenpass server address (IP:port) of the remote peer when receiving this message
-	// This value is the local Rosenpass server address when sending the message
-	RosenpassAddr string
-
-	// relay server address
-	RelaySrvAddress string
-	// RelaySrvIP is the IP the remote peer is connected to on its
-	// relay server. Used as a dial target if DNS for RelaySrvAddress
-	// fails. Zero value if the peer did not advertise an IP.
-	RelaySrvIP netip.Addr
-	// SessionID is the unique identifier of the session, used to discard old messages
-	SessionID *ICESessionID
-}
-
-func (o *OfferAnswer) hasICECredentials() bool {
-	return o.IceCredentials.UFrag != "" && o.IceCredentials.Pwd != ""
-}
-
-type Handshaker struct {
-	mu            sync.Mutex
-	log           *log.Entry
-	config        ConnConfig
-	signaler      *Signaler
-	ice           *WorkerICE
-	relay         *WorkerRelay
-	metricsStages *MetricsStages
-	// relayListener is not blocking because the listener is using a goroutine to process the messages
-	// and it will only keep the latest message if multiple offers are received in a short time
-	// this is to avoid blocking the handshaker if the listener is doing some heavy processing
-	// and also to avoid processing old offers if multiple offers are received in a short time
-	// the listener will always process the latest offer
-	relayListener *AsyncOfferListener
-	iceListener   func(remoteOfferAnswer *OfferAnswer)
-
-	// remoteICESupported tracks whether the remote peer includes ICE credentials in its offers/answers.
-	// When false, the local side skips ICE listener dispatch and suppresses ICE credentials in responses.
-	remoteICESupported atomic.Bool
-
-	// remoteOffersCh is a channel used to wait for remote credentials to proceed with the connection
-	remoteOffersCh chan OfferAnswer
-	// remoteAnswerCh is a channel used to wait for remote credentials answer (confirmation of our offer) to proceed with the connection
-	remoteAnswerCh chan OfferAnswer
-}
-
-func NewHandshaker(log *log.Entry, config ConnConfig, signaler *Signaler, ice *WorkerICE, relay *WorkerRelay, metricsStages *MetricsStages) *Handshaker {
-	h := &Handshaker{
-		log:            log,
-		config:         config,
-		signaler:       signaler,
-		ice:            ice,
-		relay:          relay,
-		metricsStages:  metricsStages,
-		remoteOffersCh: make(chan OfferAnswer),
-		remoteAnswerCh: make(chan OfferAnswer),
-	}
-	// assume remote supports ICE until we learn otherwise from received offers
-	h.remoteICESupported.Store(ice != nil)
-	return h
-}
-
-func (h *Handshaker) RemoteICESupported() bool {
-	return h.remoteICESupported.Load()
-}
-
-func (h *Handshaker) AddRelayListener(offer func(remoteOfferAnswer *OfferAnswer)) {
-	h.relayListener = NewAsyncOfferListener(offer)
-}
-
-func (h *Handshaker) AddICEListener(offer func(remoteOfferAnswer *OfferAnswer)) {
-	h.iceListener = offer
-}
-
-func (h *Handshaker) Listen(ctx context.Context) {
-	for {
-		select {
-		case remoteOfferAnswer := <-h.remoteOffersCh:
-			h.log.Infof("received offer, running version %s, remote WireGuard listen port %d, session id: %s, remote ICE supported: %t", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString(), remoteOfferAnswer.hasICECredentials())
-
-			// Record signaling received for reconnection attempts
-			if h.metricsStages != nil {
-				h.metricsStages.RecordSignalingReceived()
-			}
-
-			h.updateRemoteICEState(&remoteOfferAnswer)
-
-			if h.relayListener != nil {
-				h.relayListener.Notify(&remoteOfferAnswer)
-			}
-
-			if h.iceListener != nil && h.RemoteICESupported() {
-				h.iceListener(&remoteOfferAnswer)
-			}
-
-			if err := h.sendAnswer(); err != nil {
-				h.log.Errorf("failed to send remote offer confirmation: %s", err)
-				continue
-			}
-		case remoteOfferAnswer := <-h.remoteAnswerCh:
-			h.log.Infof("received answer, running version %s, remote WireGuard listen port %d, session id: %s, remote ICE supported: %t", remoteOfferAnswer.Version, remoteOfferAnswer.WgListenPort, remoteOfferAnswer.SessionIDString(), remoteOfferAnswer.hasICECredentials())
-
-			// Record signaling received for reconnection attempts
-			if h.metricsStages != nil {
-				h.metricsStages.RecordSignalingReceived()
-			}
-
-			h.updateRemoteICEState(&remoteOfferAnswer)
-
-			if h.relayListener != nil {
-				h.relayListener.Notify(&remoteOfferAnswer)
-			}
-
-			if h.iceListener != nil && h.RemoteICESupported() {
-				h.iceListener(&remoteOfferAnswer)
-			}
-		case <-ctx.Done():
-			h.log.Infof("stop listening for remote offers and answers")
-			return
-		}
-	}
-}
-
-func (h *Handshaker) SendOffer() error {
-	h.mu.Lock()
-	defer h.mu.Unlock()
-	return h.sendOffer()
-}
-
-// OnRemoteOffer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
-// doesn't block, discards the message if connection wasn't ready
-func (h *Handshaker) OnRemoteOffer(offer OfferAnswer) {
-	select {
-	case h.remoteOffersCh <- offer:
-		return
-	default:
-		h.log.Warnf("skipping remote offer message because receiver not ready")
-		// connection might not be ready yet to receive so we ignore the message
-		return
-	}
-}
-
-// OnRemoteAnswer handles an offer from the remote peer and returns true if the message was accepted, false otherwise
-// doesn't block, discards the message if connection wasn't ready
-func (h *Handshaker) OnRemoteAnswer(answer OfferAnswer) {
-	select {
-	case h.remoteAnswerCh <- answer:
-		return
-	default:
-		// connection might not be ready yet to receive so we ignore the message
-		h.log.Warnf("skipping remote answer message because receiver not ready")
-		return
-	}
-}
-
-// sendOffer prepares local user credentials and signals them to the remote peer
-func (h *Handshaker) sendOffer() error {
-	if !h.signaler.Ready() {
-		return ErrSignalIsNotReady
-	}
-
-	offer := h.buildOfferAnswer()
-	h.log.Debugf("sending offer with serial: %s", offer.SessionIDString())
-
-	return h.signaler.SignalOffer(offer, h.config.Key)
-}
-
-func (h *Handshaker) sendAnswer() error {
-	answer := h.buildOfferAnswer()
-	h.log.Debugf("sending answer with serial: %s", answer.SessionIDString())
-
-	return h.signaler.SignalAnswer(answer, h.config.Key)
-}
-
-func (h *Handshaker) buildOfferAnswer() OfferAnswer {
-	answer := OfferAnswer{
-		WgListenPort:    h.config.LocalWgPort,
-		Version:         version.NetbirdVersion(),
-		RosenpassPubKey: h.config.RosenpassConfig.PubKey,
-		RosenpassAddr:   h.config.RosenpassConfig.Addr,
-	}
-
-	if h.ice != nil && h.RemoteICESupported() {
-		uFrag, pwd := h.ice.GetLocalUserCredentials()
-		sid := h.ice.SessionID()
-		answer.IceCredentials = IceCredentials{uFrag, pwd}
-		answer.SessionID = &sid
-	}
-
-	if addr, ip, err := h.relay.RelayInstanceAddress(); err == nil {
-		answer.RelaySrvAddress = addr
-		answer.RelaySrvIP = ip
-	}
-
-	return answer
-}
-
-func (h *Handshaker) updateRemoteICEState(offer *OfferAnswer) {
-	hasICE := offer.hasICECredentials()
-	prev := h.remoteICESupported.Swap(hasICE)
-	if prev != hasICE {
-		if hasICE {
-			h.log.Infof("remote peer started sending ICE credentials")
-		} else {
-			h.log.Infof("remote peer stopped sending ICE credentials")
-			if h.ice != nil {
-				h.ice.Close()
-			}
-		}
-	}
-}

client/internal/peer/handshaker_listener.go

@@ -1,62 +0,0 @@
-package peer
-
-import (
-	"sync"
-)
-
-type callbackFunc func(remoteOfferAnswer *OfferAnswer)
-
-func (oa *OfferAnswer) SessionIDString() string {
-	if oa.SessionID == nil {
-		return "unknown"
-	}
-	return oa.SessionID.String()
-}
-
-type AsyncOfferListener struct {
-	fn      callbackFunc
-	running bool
-	latest  *OfferAnswer
-	mu      sync.Mutex
-}
-
-func NewAsyncOfferListener(fn callbackFunc) *AsyncOfferListener {
-	return &AsyncOfferListener{
-		fn: fn,
-	}
-}
-
-func (o *AsyncOfferListener) Notify(remoteOfferAnswer *OfferAnswer) {
-	o.mu.Lock()
-	defer o.mu.Unlock()
-
-	// Store the latest offer
-	o.latest = remoteOfferAnswer
-
-	// If already running, the running goroutine will pick up this latest value
-	if o.running {
-		return
-	}
-
-	// Start processing
-	o.running = true
-
-	// Process in a goroutine to avoid blocking the caller
-	go func(remoteOfferAnswer *OfferAnswer) {
-		for {
-			o.fn(remoteOfferAnswer)
-
-			o.mu.Lock()
-			if o.latest == nil {
-				// No more work to do
-				o.running = false
-				o.mu.Unlock()
-				return
-			}
-			remoteOfferAnswer = o.latest
-			// Clear the latest to mark it as being processed
-			o.latest = nil
-			o.mu.Unlock()
-		}
-	}(remoteOfferAnswer)
-}

client/internal/peer/handshaker_listener_test.go

@@ -1,39 +0,0 @@
-package peer
-
-import (
-	"testing"
-	"time"
-)
-
-func Test_newOfferListener(t *testing.T) {
-	dummyOfferAnswer := &OfferAnswer{}
-	runChan := make(chan struct{}, 10)
-
-	longRunningFn := func(remoteOfferAnswer *OfferAnswer) {
-		time.Sleep(1 * time.Second)
-		runChan <- struct{}{}
-	}
-
-	hl := NewAsyncOfferListener(longRunningFn)
-
-	hl.Notify(dummyOfferAnswer)
-	hl.Notify(dummyOfferAnswer)
-	hl.Notify(dummyOfferAnswer)
-
-	// Wait for exactly 2 callbacks
-	for i := 0; i < 2; i++ {
-		select {
-		case <-runChan:
-		case <-time.After(3 * time.Second):
-			t.Fatal("Timeout waiting for callback")
-		}
-	}
-
-	// Verify no additional callbacks happen
-	select {
-	case <-runChan:
-		t.Fatal("Unexpected additional callback")
-	case <-time.After(100 * time.Millisecond):
-		t.Log("Correctly received exactly 2 callbacks")
-	}
-}

client/internal/peer/ice/session_id.go

@@ -1,4 +1,4 @@
-package peer
+package ice
 
 import (
 	"crypto/rand"
@@ -9,26 +9,26 @@ import (
 
 const sessionIDSize = 5
 
-type ICESessionID string
+type SessionID string
 
-// NewICESessionID generates a new session ID for distinguishing sessions
-func NewICESessionID() (ICESessionID, error) {
+// NewSessionID generates a new session ID for distinguishing sessions
+func NewSessionID() (SessionID, error) {
 	b := make([]byte, sessionIDSize)
 	if _, err := io.ReadFull(rand.Reader, b); err != nil {
 		return "", fmt.Errorf("failed to generate session ID: %w", err)
 	}
-	return ICESessionID(hex.EncodeToString(b)), nil
+	return SessionID(hex.EncodeToString(b)), nil
 }
 
-func ICESessionIDFromBytes(b []byte) (ICESessionID, error) {
+func SessionIDFromBytes(b []byte) (SessionID, error) {
 	if len(b) != sessionIDSize {
 		return "", fmt.Errorf("invalid session ID length: %d", len(b))
 	}
-	return ICESessionID(hex.EncodeToString(b)), nil
+	return SessionID(hex.EncodeToString(b)), nil
 }
 
 // Bytes returns the raw bytes of the session ID for protobuf serialization
-func (id ICESessionID) Bytes() ([]byte, error) {
+func (id SessionID) Bytes() ([]byte, error) {
 	if len(id) == 0 {
 		return nil, fmt.Errorf("ICE session ID is empty")
 	}
@@ -42,6 +42,6 @@ func (id ICESessionID) Bytes() ([]byte, error) {
 	return b, nil
 }
 
-func (id ICESessionID) String() string {
+func (id SessionID) String() string {
 	return string(id)
 }

client/internal/peer/iface.go

@@ -1,22 +0,0 @@
-package peer
-
-import (
-	"net"
-	"net/netip"
-	"time"
-
-	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
-
-	"github.com/netbirdio/netbird/client/iface/configurer"
-	"github.com/netbirdio/netbird/client/iface/wgaddr"
-	"github.com/netbirdio/netbird/client/iface/wgproxy"
-)
-
-type WGIface interface {
-	UpdatePeer(peerKey string, allowedIps []netip.Prefix, keepAlive time.Duration, endpoint *net.UDPAddr, preSharedKey *wgtypes.Key) error
-	RemovePeer(peerKey string) error
-	GetStats() (map[string]configurer.WGStats, error)
-	GetProxy() wgproxy.Proxy
-	Address() wgaddr.Address
-	RemoveEndpointAddress(key string) error
-}

client/internal/peer/listener.go

@@ -1,11 +0,0 @@
-package peer
-
-// Listener is a callback type about the NetBird network connection state
-type Listener interface {
-	OnConnected()
-	OnDisconnected()
-	OnConnecting()
-	OnDisconnecting()
-	OnAddressChanged(string, string)
-	OnPeersListChanged(int)
-}

client/internal/peer/mailbox.go

@@ -0,0 +1,116 @@
+package peer
+
+import (
+	"sync"
+)
+
+// maxQueuedCandidates bounds the remote candidate queue; on overflow the
+// oldest candidate is dropped. Lost candidates are recovered by the next
+// offer exchange triggered by the guard.
+const maxQueuedCandidates = 128
+
+// mailbox is the coalescing inbox of the Conn event loop. Posting never
+// blocks. Per message kind either the latest value wins (offer, answer,
+// guard tick), the values queue in bounded FIFO order (candidates) or in
+// unbounded FIFO order (lifecycle and transport state changes, which are
+// low-volume and must not be lost). A new offer flushes the queued
+// candidates because they belong to the superseded session.
+type mailbox struct {
+	mu     sync.Mutex
+	closed bool
+
+	lifecycle  []event
+	transport  []event
+	offer      *evRemoteOffer
+	answer     *evRemoteAnswer
+	candidates []evRemoteCandidate
+	guardTick  bool
+
+	wake chan struct{}
+}
+
+func newMailbox() *mailbox {
+	return &mailbox{
+		wake: make(chan struct{}, 1),
+	}
+}
+
+// post stores the event and wakes the loop. It reports false if the mailbox
+// is already closed and the event was not accepted.
+func (m *mailbox) post(ev event) bool {
+	m.mu.Lock()
+	if m.closed {
+		m.mu.Unlock()
+		return false
+	}
+
+	switch e := ev.(type) {
+	case evClose:
+		m.lifecycle = append(m.lifecycle, e)
+	case evRemoteOffer:
+		m.offer = &e
+		m.candidates = nil
+	case evRemoteAnswer:
+		m.answer = &e
+	case evRemoteCandidate:
+		if len(m.candidates) >= maxQueuedCandidates {
+			m.candidates = m.candidates[1:]
+		}
+		m.candidates = append(m.candidates, e)
+	case evGuardTick:
+		m.guardTick = true
+	default:
+		m.transport = append(m.transport, ev)
+	}
+	m.mu.Unlock()
+
+	select {
+	case m.wake <- struct{}{}:
+	default:
+	}
+	return true
+}
+
+// drain returns the pending events in processing order: lifecycle first,
+// then transport state changes, the coalesced offer and answer, the queued
+// candidates and finally the guard tick.
+func (m *mailbox) drain() []event {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	return m.drainLocked()
+}
+
+// closeAndDrain marks the mailbox closed so further posts are rejected and
+// returns the events that were still pending.
+func (m *mailbox) closeAndDrain() []event {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	m.closed = true
+	return m.drainLocked()
+}
+
+func (m *mailbox) drainLocked() []event {
+	evs := make([]event, 0, len(m.lifecycle)+len(m.transport)+len(m.candidates)+3)
+	evs = append(evs, m.lifecycle...)
+	evs = append(evs, m.transport...)
+	if m.offer != nil {
+		evs = append(evs, *m.offer)
+	}
+	if m.answer != nil {
+		evs = append(evs, *m.answer)
+	}
+	for _, c := range m.candidates {
+		evs = append(evs, c)
+	}
+	if m.guardTick {
+		evs = append(evs, evGuardTick{})
+	}
+
+	m.lifecycle = nil
+	m.transport = nil
+	m.offer = nil
+	m.answer = nil
+	m.candidates = nil
+	m.guardTick = false
+	return evs
+}

client/internal/peer/mailbox_test.go

@@ -0,0 +1,128 @@
+package peer
+
+import (
+	"testing"
+
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestMailbox_OfferCoalescing(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{WgListenPort: 1}}))
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{WgListenPort: 2}}))
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{WgListenPort: 3}}))
+
+	evs := mb.drain()
+	require.Len(t, evs, 1, "consecutive offers must coalesce to a single event")
+	offer, ok := evs[0].(evRemoteOffer)
+	require.True(t, ok, "coalesced event must be an offer")
+	assert.Equal(t, 3, offer.offer.WgListenPort, "the newest offer must win")
+}
+
+func TestMailbox_OfferFlushesCandidates(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evRemoteCandidate{}))
+	require.True(t, mb.post(evRemoteCandidate{}))
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{}}))
+
+	evs := mb.drain()
+	require.Len(t, evs, 1, "candidates of the superseded session must be flushed")
+	_, ok := evs[0].(evRemoteOffer)
+	assert.True(t, ok, "only the offer must remain after the flush")
+}
+
+func TestMailbox_CandidatesKeepOrderAfterOffer(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{}}))
+	require.True(t, mb.post(evRemoteCandidate{haRoutes: nil}))
+	require.True(t, mb.post(evRemoteCandidate{haRoutes: nil}))
+
+	evs := mb.drain()
+	require.Len(t, evs, 3)
+	_, ok := evs[0].(evRemoteOffer)
+	assert.True(t, ok, "offer must be processed before the candidates")
+	for _, ev := range evs[1:] {
+		_, ok := ev.(evRemoteCandidate)
+		assert.True(t, ok, "candidates posted after the offer must survive")
+	}
+}
+
+func TestMailbox_CandidateQueueBounded(t *testing.T) {
+	mb := newMailbox()
+
+	for i := 0; i < maxQueuedCandidates+10; i++ {
+		require.True(t, mb.post(evRemoteCandidate{}))
+	}
+
+	evs := mb.drain()
+	assert.Len(t, evs, maxQueuedCandidates, "candidate queue must stay bounded")
+}
+
+func TestMailbox_DrainOrder(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evGuardTick{}))
+	require.True(t, mb.post(evRemoteAnswer{answer: signaling.OfferAnswer{}}))
+	require.True(t, mb.post(evRemoteOffer{offer: signaling.OfferAnswer{}}))
+	require.True(t, mb.post(evRelayDown{}))
+	require.True(t, mb.post(evICEDown{sessionChanged: true}))
+	require.True(t, mb.post(evClose{}))
+
+	evs := mb.drain()
+	require.Len(t, evs, 6)
+
+	_, ok := evs[0].(evClose)
+	assert.True(t, ok, "lifecycle events must come first")
+	_, ok = evs[1].(evRelayDown)
+	assert.True(t, ok, "transport events must keep FIFO order")
+	_, ok = evs[2].(evICEDown)
+	assert.True(t, ok, "transport events must keep FIFO order")
+	_, ok = evs[3].(evRemoteOffer)
+	assert.True(t, ok, "offer must come after transport events")
+	_, ok = evs[4].(evRemoteAnswer)
+	assert.True(t, ok, "answer must come after the offer")
+	_, ok = evs[5].(evGuardTick)
+	assert.True(t, ok, "guard tick must come last")
+}
+
+func TestMailbox_GuardTickCoalesced(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evGuardTick{}))
+	require.True(t, mb.post(evGuardTick{}))
+	require.True(t, mb.post(evGuardTick{}))
+
+	evs := mb.drain()
+	assert.Len(t, evs, 1, "guard ticks must coalesce to a single event")
+}
+
+func TestMailbox_PostAfterCloseRejected(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evRelayDown{}))
+	leftovers := mb.closeAndDrain()
+	assert.Len(t, leftovers, 1, "pending events must be returned on close")
+
+	assert.False(t, mb.post(evRelayDown{}), "posts must be rejected after close")
+	assert.Empty(t, mb.drain(), "no events must remain after close")
+}
+
+func TestMailbox_WakeSignal(t *testing.T) {
+	mb := newMailbox()
+
+	require.True(t, mb.post(evRelayDown{}))
+	require.True(t, mb.post(evGuardTick{}))
+
+	select {
+	case <-mb.wake:
+	default:
+		t.Fatal("wake signal must be pending after posts")
+	}
+
+	assert.Len(t, mb.drain(), 2, "a single wake must deliver all pending events")
+}

client/internal/peer/metricsstages/metrics_saver.go

@@ -1,4 +1,4 @@
-package peer
+package metricsstages
 
 import (
 	"sync"

client/internal/peer/metricsstages/metrics_saver_test.go

@@ -1,4 +1,4 @@
-package peer
+package metricsstages
 
 import (
 	"testing"

client/internal/peer/signaling/handshaker.go

@@ -0,0 +1,189 @@
+package signaling
+
+import (
+	"errors"
+	"net/netip"
+	"sync"
+	"sync/atomic"
+
+	log "github.com/sirupsen/logrus"
+
+	icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
+	relayClient "github.com/netbirdio/netbird/shared/relay/client"
+	"github.com/netbirdio/netbird/version"
+)
+
+var (
+	ErrSignalIsNotReady = errors.New("signal is not ready")
+)
+
+// IceCredentials ICE protocol credentials struct
+type IceCredentials struct {
+	UFrag string
+	Pwd   string
+}
+
+// OfferAnswer represents a session establishment offer or answer
+type OfferAnswer struct {
+	IceCredentials IceCredentials
+	// WgListenPort is a remote WireGuard listen port.
+	// This field is used when establishing a direct WireGuard connection without any proxy.
+	// We can set the remote peer's endpoint with this port.
+	WgListenPort int
+
+	// Version of NetBird Agent
+	Version string
+	// RosenpassPubKey is the Rosenpass public key of the remote peer when receiving this message
+	// This value is the local Rosenpass server public key when sending the message
+	RosenpassPubKey []byte
+	// RosenpassAddr is the Rosenpass server address (IP:port) of the remote peer when receiving this message
+	// This value is the local Rosenpass server address when sending the message
+	RosenpassAddr string
+
+	// relay server address
+	RelaySrvAddress string
+	// RelaySrvIP is the IP the remote peer is connected to on its
+	// relay server. Used as a dial target if DNS for RelaySrvAddress
+	// fails. Zero value if the peer did not advertise an IP.
+	RelaySrvIP netip.Addr
+	// SessionID is the unique identifier of the session, used to discard old messages
+	SessionID *icemaker.SessionID
+}
+
+func (o *OfferAnswer) HasICECredentials() bool {
+	return o.IceCredentials.UFrag != "" && o.IceCredentials.Pwd != ""
+}
+
+func (o *OfferAnswer) SessionIDString() string {
+	if o.SessionID == nil {
+		return "unknown"
+	}
+	return o.SessionID.String()
+}
+
+// Config carries the peer-specific values the Handshaker embeds into offers
+// and answers.
+type Config struct {
+	Key             string
+	LocalWgPort     int
+	RosenpassPubKey []byte
+	RosenpassAddr   string
+}
+
+// Credentials are the local ICE credentials and session id the Handshaker embeds in offers.
+type Credentials struct {
+	UFrag     string
+	Pwd       string
+	SessionID icemaker.SessionID
+}
+
+// ICEWorker is the subset of the ICE worker the Handshaker needs to build offers.
+type ICEWorker interface {
+	Credentials() Credentials
+	Close()
+}
+
+// Handshaker keeps the signaling protocol logic: building and sending offers
+// and answers and tracking whether the remote peer supports ICE. Incoming
+// message processing is driven by the Conn event loop.
+type Handshaker struct {
+	mu           sync.Mutex
+	log          *log.Entry
+	config       Config
+	signaler     *Signaler
+	ice          ICEWorker
+	relayManager *relayClient.Manager
+
+	// remoteICESupported tracks whether the remote peer includes ICE credentials in its offers/answers.
+	// When false, the local side skips ICE dispatch and suppresses ICE credentials in responses.
+	remoteICESupported atomic.Bool
+}
+
+func NewHandshaker(log *log.Entry, config Config, signaler *Signaler, ice ICEWorker, relayManager *relayClient.Manager) *Handshaker {
+	h := &Handshaker{
+		log:          log,
+		config:       config,
+		signaler:     signaler,
+		ice:          ice,
+		relayManager: relayManager,
+	}
+	// assume remote supports ICE until we learn otherwise from received offers
+	h.remoteICESupported.Store(ice != nil)
+	return h
+}
+
+func (h *Handshaker) RemoteICESupported() bool {
+	return h.remoteICESupported.Load()
+}
+
+func (h *Handshaker) SendOffer() error {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	return h.sendOffer()
+}
+
+func (h *Handshaker) SendAnswer() error {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+	return h.sendAnswer()
+}
+
+// sendOffer prepares local user credentials and signals them to the remote peer
+func (h *Handshaker) sendOffer() error {
+	if !h.signaler.Ready() {
+		return ErrSignalIsNotReady
+	}
+
+	offer := h.buildOfferAnswer()
+	h.log.Debugf("sending offer with serial: %s", offer.SessionIDString())
+
+	return h.signaler.SignalOffer(offer, h.config.Key)
+}
+
+func (h *Handshaker) sendAnswer() error {
+	answer := h.buildOfferAnswer()
+	h.log.Debugf("sending answer with serial: %s", answer.SessionIDString())
+
+	return h.signaler.SignalAnswer(answer, h.config.Key)
+}
+
+func (h *Handshaker) buildOfferAnswer() OfferAnswer {
+	answer := OfferAnswer{
+		WgListenPort:    h.config.LocalWgPort,
+		Version:         version.NetbirdVersion(),
+		RosenpassPubKey: h.config.RosenpassPubKey,
+		RosenpassAddr:   h.config.RosenpassAddr,
+	}
+
+	if h.ice != nil && h.RemoteICESupported() {
+		creds := h.ice.Credentials()
+		answer.IceCredentials = IceCredentials{creds.UFrag, creds.Pwd}
+		sid := creds.SessionID
+		answer.SessionID = &sid
+	}
+
+	if addr, ip, err := h.relayManager.RelayInstanceAddress(); err == nil {
+		answer.RelaySrvAddress = addr
+		answer.RelaySrvIP = ip
+	}
+
+	return answer
+}
+
+// UpdateRemoteICEState refreshes the remote ICE support flag from a received
+// offer or answer and closes the ICE worker when the remote peer stopped
+// sending ICE credentials. Runs on the Conn event loop.
+func (h *Handshaker) UpdateRemoteICEState(offer *OfferAnswer) {
+	hasICE := offer.HasICECredentials()
+	prev := h.remoteICESupported.Swap(hasICE)
+	if prev != hasICE {
+		if hasICE {
+			h.log.Infof("remote peer started sending ICE credentials")
+		} else {
+			h.log.Infof("remote peer stopped sending ICE credentials")
+			if h.ice != nil {
+				h.ice.Close()
+			}
+		}
+	}
+}

client/internal/peer/signaling/signaler.go

@@ -1,4 +1,4 @@
-package peer
+package signaling
 
 import (
 	"github.com/pion/ice/v4"

client/internal/peer/state_dump/state_dump.go

@@ -1,4 +1,4 @@
-package peer
+package state_dump
 
 import (
 	"context"
@@ -6,11 +6,13 @@ import (
 	"time"
 
 	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/peer/status"
 )
 
-type stateDump struct {
+type StateDump struct {
 	log    *log.Entry
-	status *Status
+	status *status.Recorder
 	key    string
 
 	sentOffer       int
@@ -26,15 +28,15 @@ type stateDump struct {
 	mu sync.Mutex
 }
 
-func newStateDump(key string, log *log.Entry, statusRecorder *Status) *stateDump {
-	return &stateDump{
+func NewStateDump(key string, log *log.Entry, statusRecorder *status.Recorder) *StateDump {
+	return &StateDump{
 		log:    log,
 		status: statusRecorder,
 		key:    key,
 	}
 }
 
-func (s *stateDump) Start(ctx context.Context) {
+func (s *StateDump) Start(ctx context.Context) {
 	ticker := time.NewTicker(10 * time.Minute)
 	defer ticker.Stop()
 
@@ -48,25 +50,25 @@ func (s *stateDump) Start(ctx context.Context) {
 	}
 }
 
-func (s *stateDump) RemoteOffer() {
+func (s *StateDump) RemoteOffer() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 	s.remoteOffer++
 }
 
-func (s *stateDump) RemoteCandidate() {
+func (s *StateDump) RemoteCandidate() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 	s.remoteCandidate++
 }
 
-func (s *stateDump) SendOffer() {
+func (s *StateDump) SendOffer() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 	s.sentOffer++
 }
 
-func (s *stateDump) dumpState() {
+func (s *StateDump) dumpState() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
@@ -80,41 +82,41 @@ func (s *stateDump) dumpState() {
 		status, s.sentOffer, s.remoteOffer, s.remoteAnswer, s.remoteCandidate, s.p2pConnected, s.switchToRelay, s.wgCheckSuccess, s.relayConnected, s.localProxies)
 }
 
-func (s *stateDump) RemoteAnswer() {
+func (s *StateDump) RemoteAnswer() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 	s.remoteAnswer++
 }
 
-func (s *stateDump) P2PConnected() {
+func (s *StateDump) P2PConnected() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	s.p2pConnected++
 }
 
-func (s *stateDump) SwitchToRelay() {
+func (s *StateDump) SwitchToRelay() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	s.switchToRelay++
 }
 
-func (s *stateDump) WGcheckSuccess() {
+func (s *StateDump) WGcheckSuccess() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	s.wgCheckSuccess++
 }
 
-func (s *stateDump) RelayConnected() {
+func (s *StateDump) RelayConnected() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	s.relayConnected++
 }
 
-func (s *stateDump) NewLocalProxy() {
+func (s *StateDump) NewLocalProxy() {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 

client/internal/peer/status/conn_status.go

@@ -0,0 +1,31 @@
+package status
+
+import (
+	log "github.com/sirupsen/logrus"
+)
+
+const (
+	// StatusIdle indicate the peer is in disconnected state
+	StatusIdle ConnStatus = iota
+	// StatusConnecting indicate the peer is in connecting state
+	StatusConnecting
+	// StatusConnected indicate the peer is in connected state
+	StatusConnected
+)
+
+// ConnStatus describe the status of a peer's connection
+type ConnStatus int32
+
+func (s ConnStatus) String() string {
+	switch s {
+	case StatusConnecting:
+		return "Connecting"
+	case StatusConnected:
+		return "Connected"
+	case StatusIdle:
+		return "Idle"
+	default:
+		log.Errorf("unknown status: %d", s)
+		return "INVALID_PEER_CONNECTION_STATUS"
+	}
+}

client/internal/peer/status/conn_status_test.go

@@ -1,4 +1,4 @@
-package peer
+package status
 
 import (
 	"testing"

client/internal/peer/status/events.go

@@ -0,0 +1,48 @@
+package status
+
+import (
+	"slices"
+	"sync"
+
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+type EventQueue struct {
+	maxSize int
+	events  []*proto.SystemEvent
+	mutex   sync.RWMutex
+}
+
+func NewEventQueue(size int) *EventQueue {
+	return &EventQueue{
+		maxSize: size,
+		events:  make([]*proto.SystemEvent, 0, size),
+	}
+}
+
+func (q *EventQueue) Add(event *proto.SystemEvent) {
+	q.mutex.Lock()
+	defer q.mutex.Unlock()
+
+	q.events = append(q.events, event)
+
+	if len(q.events) > q.maxSize {
+		q.events = q.events[len(q.events)-q.maxSize:]
+	}
+}
+
+func (q *EventQueue) GetAll() []*proto.SystemEvent {
+	q.mutex.RLock()
+	defer q.mutex.RUnlock()
+
+	return slices.Clone(q.events)
+}
+
+type EventSubscription struct {
+	id     string
+	events chan *proto.SystemEvent
+}
+
+func (s *EventSubscription) Events() <-chan *proto.SystemEvent {
+	return s.events
+}

client/internal/peer/status/full_status.go

@@ -0,0 +1,122 @@
+package status
+
+import (
+	"golang.org/x/exp/maps"
+	"google.golang.org/protobuf/types/known/durationpb"
+	"google.golang.org/protobuf/types/known/timestamppb"
+
+	"github.com/netbirdio/netbird/client/internal/relay"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+// FullStatus contains the full state held by the Recorder instance
+type FullStatus struct {
+	Peers                 []State
+	ManagementState       ManagementState
+	SignalState           SignalState
+	LocalPeerState        LocalPeerState
+	RosenpassState        RosenpassState
+	Relays                []relay.ProbeResult
+	NSGroupStates         []NSGroupState
+	NumOfForwardingRules  int
+	LazyConnectionEnabled bool
+	Events                []*proto.SystemEvent
+}
+
+// ToProto converts FullStatus to proto.FullStatus.
+func (fs FullStatus) ToProto() *proto.FullStatus {
+	pbFullStatus := proto.FullStatus{
+		ManagementState: &proto.ManagementState{},
+		SignalState:     &proto.SignalState{},
+		LocalPeerState:  &proto.LocalPeerState{},
+		Peers:           []*proto.PeerState{},
+	}
+
+	pbFullStatus.ManagementState.URL = fs.ManagementState.URL
+	pbFullStatus.ManagementState.Connected = fs.ManagementState.Connected
+	if err := fs.ManagementState.Error; err != nil {
+		pbFullStatus.ManagementState.Error = err.Error()
+	}
+
+	pbFullStatus.SignalState.URL = fs.SignalState.URL
+	pbFullStatus.SignalState.Connected = fs.SignalState.Connected
+	if err := fs.SignalState.Error; err != nil {
+		pbFullStatus.SignalState.Error = err.Error()
+	}
+
+	pbFullStatus.LocalPeerState.IP = fs.LocalPeerState.IP
+	pbFullStatus.LocalPeerState.Ipv6 = fs.LocalPeerState.IPv6
+	pbFullStatus.LocalPeerState.PubKey = fs.LocalPeerState.PubKey
+	pbFullStatus.LocalPeerState.KernelInterface = fs.LocalPeerState.KernelInterface
+	pbFullStatus.LocalPeerState.Fqdn = fs.LocalPeerState.FQDN
+	pbFullStatus.LocalPeerState.WgPort = int32(fs.LocalPeerState.WgPort)
+	pbFullStatus.LocalPeerState.RosenpassPermissive = fs.RosenpassState.Permissive
+	pbFullStatus.LocalPeerState.RosenpassEnabled = fs.RosenpassState.Enabled
+	pbFullStatus.NumberOfForwardingRules = int32(fs.NumOfForwardingRules)
+	pbFullStatus.LazyConnectionEnabled = fs.LazyConnectionEnabled
+
+	pbFullStatus.LocalPeerState.Networks = maps.Keys(fs.LocalPeerState.Routes)
+
+	for _, peerState := range fs.Peers {
+		networks := maps.Keys(peerState.GetRoutes())
+
+		pbPeerState := &proto.PeerState{
+			IP:                         peerState.IP,
+			Ipv6:                       peerState.IPv6,
+			PubKey:                     peerState.PubKey,
+			ConnStatus:                 peerState.ConnStatus.String(),
+			ConnStatusUpdate:           timestamppb.New(peerState.ConnStatusUpdate),
+			Relayed:                    peerState.Relayed,
+			LocalIceCandidateType:      peerState.LocalIceCandidateType,
+			RemoteIceCandidateType:     peerState.RemoteIceCandidateType,
+			LocalIceCandidateEndpoint:  peerState.LocalIceCandidateEndpoint,
+			RemoteIceCandidateEndpoint: peerState.RemoteIceCandidateEndpoint,
+			RelayAddress:               peerState.RelayServerAddress,
+			Fqdn:                       peerState.FQDN,
+			LastWireguardHandshake:     timestamppb.New(peerState.LastWireguardHandshake),
+			BytesRx:                    peerState.BytesRx,
+			BytesTx:                    peerState.BytesTx,
+			RosenpassEnabled:           peerState.RosenpassEnabled,
+			Networks:                   networks,
+			Latency:                    durationpb.New(peerState.Latency),
+			SshHostKey:                 peerState.SSHHostKey,
+		}
+		pbFullStatus.Peers = append(pbFullStatus.Peers, pbPeerState)
+	}
+
+	for _, relayState := range fs.Relays {
+		pbRelayState := &proto.RelayState{
+			URI:       relayState.URI,
+			Available: relayState.Err == nil,
+			Transport: relayState.Transport,
+		}
+		if err := relayState.Err; err != nil {
+			pbRelayState.Error = err.Error()
+		}
+		pbFullStatus.Relays = append(pbFullStatus.Relays, pbRelayState)
+	}
+
+	for _, dnsState := range fs.NSGroupStates {
+		var err string
+		if dnsState.Error != nil {
+			err = dnsState.Error.Error()
+		}
+
+		var servers []string
+		for _, server := range dnsState.Servers {
+			servers = append(servers, server.String())
+		}
+
+		pbDnsState := &proto.NSGroupState{
+			Servers: servers,
+			Domains: dnsState.Domains,
+			Enabled: dnsState.Enabled,
+			Error:   err,
+		}
+		pbFullStatus.DnsServers = append(pbFullStatus.DnsServers, pbDnsState)
+	}
+
+	pbFullStatus.Events = fs.Events
+
+	return &pbFullStatus
+}

client/internal/peer/status/notifier.go

@@ -1,4 +1,4 @@
-package peer
+package status
 
 import (
 	"sync"
@@ -11,6 +11,16 @@ const (
 	stateDisconnecting
 )
 
+// Listener is a callback type about the NetBird network connection state
+type Listener interface {
+	OnConnected()
+	OnDisconnected()
+	OnConnecting()
+	OnDisconnecting()
+	OnAddressChanged(string, string)
+	OnPeersListChanged(int)
+}
+
 type notifier struct {
 	serverStateLock    sync.Mutex
 	listenersLock      sync.Mutex

client/internal/peer/status/notifier_test.go

@@ -1,4 +1,4 @@
-package peer
+package status
 
 import (
 	"sync"

client/internal/peer/status/peer_state.go

@@ -0,0 +1,63 @@
+package status
+
+import (
+	"sync"
+	"time"
+
+	"golang.org/x/exp/maps"
+)
+
+// State contains the latest state of a peer
+type State struct {
+	Mux                        *sync.RWMutex
+	IP                         string
+	IPv6                       string
+	PubKey                     string
+	FQDN                       string
+	ConnStatus                 ConnStatus
+	ConnStatusUpdate           time.Time
+	Relayed                    bool
+	LocalIceCandidateType      string
+	RemoteIceCandidateType     string
+	LocalIceCandidateEndpoint  string
+	RemoteIceCandidateEndpoint string
+	RelayServerAddress         string
+	LastWireguardHandshake     time.Time
+	BytesTx                    int64
+	BytesRx                    int64
+	Latency                    time.Duration
+	RosenpassEnabled           bool
+	SSHHostKey                 []byte
+	routes                     map[string]struct{}
+}
+
+// AddRoute add a single route to routes map
+func (s *State) AddRoute(network string) {
+	s.Mux.Lock()
+	defer s.Mux.Unlock()
+	if s.routes == nil {
+		s.routes = make(map[string]struct{})
+	}
+	s.routes[network] = struct{}{}
+}
+
+// SetRoutes set state routes
+func (s *State) SetRoutes(routes map[string]struct{}) {
+	s.Mux.Lock()
+	defer s.Mux.Unlock()
+	s.routes = routes
+}
+
+// DeleteRoute removes a route from the network amp
+func (s *State) DeleteRoute(network string) {
+	s.Mux.Lock()
+	defer s.Mux.Unlock()
+	delete(s.routes, network)
+}
+
+// GetRoutes return routes map
+func (s *State) GetRoutes() map[string]struct{} {
+	s.Mux.RLock()
+	defer s.Mux.RUnlock()
+	return maps.Clone(s.routes)
+}

client/internal/peer/status/recorder_test.go

@@ -1,4 +1,4 @@
-package peer
+package status
 
 import (
 	"context"
@@ -314,3 +314,39 @@ func TestGetFullStatus(t *testing.T) {
 	assert.Equal(t, signalState, fullStatus.SignalState, "signal status should be equal")
 	assert.ElementsMatch(t, []State{peerState1, peerState2}, fullStatus.Peers, "peers states should match")
 }
+
+// notified reports whether a state-change tick is pending on ch, draining it.
+func notified(ch <-chan struct{}) bool {
+	select {
+	case <-ch:
+		return true
+	default:
+		return false
+	}
+}
+
+func TestMarkServerStateDoesNotNotifyWhenUnchanged(t *testing.T) {
+	status := NewRecorder("https://mgm")
+	_, ch := status.SubscribeToStateChanges()
+
+	// First transition is a real change and must notify.
+	status.MarkManagementConnected()
+	require.True(t, notified(ch), "first connect should notify")
+
+	// Re-marking the same state must not notify again.
+	status.MarkManagementConnected()
+	assert.False(t, notified(ch), "redundant connect should not notify")
+
+	// Same for signal.
+	status.MarkSignalConnected()
+	require.True(t, notified(ch), "first signal connect should notify")
+	status.MarkSignalConnected()
+	assert.False(t, notified(ch), "redundant signal connect should not notify")
+
+	// A genuine change (disconnect with an error) notifies again.
+	err := errors.New("boom")
+	status.MarkManagementDisconnected(err)
+	require.True(t, notified(ch), "disconnect should notify")
+	status.MarkManagementDisconnected(err)
+	assert.False(t, notified(ch), "redundant disconnect should not notify")
+}

client/internal/peer/status/route.go

@@ -1,4 +1,4 @@
-package peer
+package status
 
 import (
 	"net/netip"

client/internal/peer/status_alias.go

@@ -0,0 +1,36 @@
+package peer
+
+import "github.com/netbirdio/netbird/client/internal/peer/status"
+
+// Transitional aliases re-exporting the peer status recorder from its own
+// package. Callers are being migrated to reference the status package
+// directly; these aliases will be removed once the migration completes.
+type (
+	Status                   = status.Recorder
+	State                    = status.State
+	ConnStatus               = status.ConnStatus
+	FullStatus               = status.FullStatus
+	RouterState              = status.RouterState
+	LocalPeerState           = status.LocalPeerState
+	SignalState              = status.SignalState
+	ManagementState          = status.ManagementState
+	RosenpassState           = status.RosenpassState
+	NSGroupState             = status.NSGroupState
+	ResolvedDomainInfo       = status.ResolvedDomainInfo
+	StatusChangeSubscription = status.StatusChangeSubscription
+	EventQueue               = status.EventQueue
+	EventSubscription        = status.EventSubscription
+	WGIfaceStatus            = status.WGIfaceStatus
+	Listener                 = status.Listener
+	EventListener            = status.EventListener
+)
+
+const (
+	StatusIdle       = status.StatusIdle
+	StatusConnecting = status.StatusConnecting
+	StatusConnected  = status.StatusConnected
+)
+
+var (
+	NewRecorder = status.NewRecorder
+)

client/internal/peer/wg_watcher/wg_watcher.go

@@ -1,4 +1,4 @@
-package peer
+package wg_watcher
 
 import (
 	"context"
@@ -9,6 +9,7 @@ import (
 	log "github.com/sirupsen/logrus"
 
 	"github.com/netbirdio/netbird/client/iface/configurer"
+	"github.com/netbirdio/netbird/client/internal/peer/state_dump"
 )
 
 const (
@@ -28,7 +29,7 @@ type WGWatcher struct {
 	log           *log.Entry
 	wgIfaceStater WGInterfaceStater
 	peerKey       string
-	stateDump     *stateDump
+	stateDump     *state_dump.StateDump
 
 	enabled   bool
 	muEnabled sync.Mutex
@@ -38,7 +39,7 @@ type WGWatcher struct {
 	resetCh chan struct{}
 }
 
-func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey string, stateDump *stateDump) *WGWatcher {
+func NewWGWatcher(log *log.Entry, wgIfaceStater WGInterfaceStater, peerKey string, stateDump *state_dump.StateDump) *WGWatcher {
 	return &WGWatcher{
 		log:           log,
 		wgIfaceStater: wgIfaceStater,
@@ -71,9 +72,11 @@ func (w *WGWatcher) PrepareInitialHandshake() (ok bool) {
 
 // EnableWgWatcher runs the WireGuard watcher loop using the handshake baseline captured by
 // PrepareInitialHandshake. The watcher runs until ctx is cancelled. Caller is responsible
-// for context lifecycle management.
-func (w *WGWatcher) EnableWgWatcher(ctx context.Context, enabledTime time.Time, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time)) {
-	w.periodicHandshakeCheck(ctx, onDisconnectedFn, onHandshakeSuccessFn, enabledTime, w.initialHandshake)
+// for context lifecycle management. onHandshakeSuccessFn is called only for the first
+// handshake observed by this run, onCheckSuccessFn for every check that observed a fresh
+// handshake, including the first.
+func (w *WGWatcher) EnableWgWatcher(ctx context.Context, enabledTime time.Time, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time), onCheckSuccessFn func()) {
+	w.periodicHandshakeCheck(ctx, onDisconnectedFn, onHandshakeSuccessFn, onCheckSuccessFn, enabledTime, w.initialHandshake)
 
 	w.muEnabled.Lock()
 	w.enabled = false
@@ -90,7 +93,7 @@ func (w *WGWatcher) Reset() {
 }
 
 // wgStateCheck help to check the state of the WireGuard handshake and relay connection
-func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time), enabledTime time.Time, initialHandshake time.Time) {
+func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, onDisconnectedFn func(), onHandshakeSuccessFn func(when time.Time), onCheckSuccessFn func(), enabledTime time.Time, initialHandshake time.Time) {
 	w.log.Infof("WireGuard watcher started")
 
 	timer := time.NewTimer(wgHandshakeOvertime)
@@ -117,6 +120,10 @@ func (w *WGWatcher) periodicHandshakeCheck(ctx context.Context, onDisconnectedFn
 				}
 			}
 
+			if onCheckSuccessFn != nil && ctx.Err() == nil {
+				onCheckSuccessFn()
+			}
+
 			lastHandshake = *handshake
 
 			resetTime := time.Until(handshake.Add(checkPeriod))

client/internal/peer/wg_watcher/wg_watcher_test.go

@@ -0,0 +1,179 @@
+package wg_watcher
+
+import (
+	"context"
+	"sync"
+	"testing"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/iface/configurer"
+	"github.com/netbirdio/netbird/client/internal/peer/state_dump"
+	"github.com/netbirdio/netbird/client/internal/peer/status"
+)
+
+type MocWgIface struct {
+	stop bool
+}
+
+func (m *MocWgIface) GetStats() (map[string]configurer.WGStats, error) {
+	return map[string]configurer.WGStats{}, nil
+}
+
+func (m *MocWgIface) disconnect() {
+	m.stop = true
+}
+
+type mockHandshakeStats struct {
+	mu        sync.Mutex
+	handshake time.Time
+}
+
+func (m *mockHandshakeStats) GetStats() (map[string]configurer.WGStats, error) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	return map[string]configurer.WGStats{"": {LastHandshake: m.handshake}}, nil
+}
+
+func (m *mockHandshakeStats) advance() {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	m.handshake = time.Now()
+}
+
+// TestWGWatcher_CheckSuccessCallback: onCheckSuccessFn must fire for a fresh
+// handshake even when the watcher started with an existing handshake baseline,
+// the case where onHandshakeSuccessFn stays silent.
+func TestWGWatcher_CheckSuccessCallback(t *testing.T) {
+	// checkPeriod bounds how stale a handshake may be before the watcher treats it
+	// as a suspended-machine timeout. The first check fires after wgHandshakeOvertime,
+	// so keep checkPeriod well above any scheduling jitter to avoid a false timeout
+	// converting the expected success into a disconnect on a loaded runner.
+	checkPeriod = 1 * time.Minute
+	wgHandshakeOvertime = 1 * time.Second
+
+	mlog := log.WithField("peer", "tet")
+	// Use an old baseline so advance() yields a strictly newer handshake even on
+	// platforms with coarse clock resolution (Windows), where two time.Now() calls
+	// microseconds apart can return the same instant and read as a timed-out handshake.
+	stats := &mockHandshakeStats{handshake: time.Now().Add(-time.Hour)}
+	watcher := NewWGWatcher(mlog, stats, "", state_dump.NewStateDump("peer", mlog, &status.Recorder{}))
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	require.True(t, watcher.PrepareInitialHandshake())
+
+	firstHandshake := make(chan struct{}, 1)
+	checkSuccess := make(chan struct{}, 1)
+	watcherDone := make(chan struct{})
+	go func() {
+		defer close(watcherDone)
+		watcher.EnableWgWatcher(ctx, time.Now(), func() {}, func(when time.Time) {
+			firstHandshake <- struct{}{}
+		}, func() {
+			select {
+			case checkSuccess <- struct{}{}:
+			default:
+			}
+		})
+	}()
+
+	stats.advance()
+
+	select {
+	case <-checkSuccess:
+	case <-time.After(10 * time.Second):
+		t.Errorf("timeout waiting for check success callback")
+	}
+
+	select {
+	case <-firstHandshake:
+		t.Errorf("first-handshake callback must not fire for a non-zero baseline")
+	default:
+	}
+
+	// Wait for the watcher goroutine to exit so it cannot race with other
+	// tests mutating the package-level check timing variables.
+	cancel()
+	<-watcherDone
+}
+
+func TestWGWatcher_EnableWgWatcher(t *testing.T) {
+	checkPeriod = 5 * time.Second
+	wgHandshakeOvertime = 1 * time.Second
+
+	mlog := log.WithField("peer", "tet")
+	mocWgIface := &MocWgIface{}
+	watcher := NewWGWatcher(mlog, mocWgIface, "", state_dump.NewStateDump("peer", mlog, &status.Recorder{}))
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	ok := watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should not be enabled yet")
+
+	onDisconnected := make(chan struct{}, 1)
+	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
+		mlog.Infof("onDisconnectedFn")
+		onDisconnected <- struct{}{}
+	}, func(when time.Time) {
+		mlog.Infof("onHandshakeSuccess: %v", when)
+	}, nil)
+
+	// wait for initial reading
+	time.Sleep(2 * time.Second)
+	mocWgIface.disconnect()
+
+	select {
+	case <-onDisconnected:
+	case <-time.After(10 * time.Second):
+		t.Errorf("timeout")
+	}
+}
+
+func TestWGWatcher_ReEnable(t *testing.T) {
+	checkPeriod = 5 * time.Second
+	wgHandshakeOvertime = 1 * time.Second
+
+	mlog := log.WithField("peer", "tet")
+	mocWgIface := &MocWgIface{}
+	watcher := NewWGWatcher(mlog, mocWgIface, "", state_dump.NewStateDump("peer", mlog, &status.Recorder{}))
+
+	ctx, cancel := context.WithCancel(context.Background())
+	ok := watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should not be enabled yet")
+
+	wg := &sync.WaitGroup{}
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		watcher.EnableWgWatcher(ctx, time.Now(), func() {}, func(when time.Time) {}, nil)
+	}()
+	cancel()
+
+	wg.Wait()
+
+	// Re-enable with a new context
+	ctx, cancel = context.WithCancel(context.Background())
+	defer cancel()
+
+	ok = watcher.PrepareInitialHandshake()
+	require.True(t, ok, "watcher should be re-enabled after the previous run stopped")
+
+	onDisconnected := make(chan struct{}, 1)
+	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
+		onDisconnected <- struct{}{}
+	}, func(when time.Time) {}, nil)
+
+	time.Sleep(2 * time.Second)
+	mocWgIface.disconnect()
+
+	select {
+	case <-onDisconnected:
+	case <-time.After(10 * time.Second):
+		t.Errorf("timeout")
+	}
+}

client/internal/peer/wg_watcher_test.go

@@ -1,102 +0,0 @@
-package peer
-
-import (
-	"context"
-	"sync"
-	"testing"
-	"time"
-
-	log "github.com/sirupsen/logrus"
-	"github.com/stretchr/testify/require"
-
-	"github.com/netbirdio/netbird/client/iface/configurer"
-)
-
-type MocWgIface struct {
-	stop bool
-}
-
-func (m *MocWgIface) GetStats() (map[string]configurer.WGStats, error) {
-	return map[string]configurer.WGStats{}, nil
-}
-
-func (m *MocWgIface) disconnect() {
-	m.stop = true
-}
-
-func TestWGWatcher_EnableWgWatcher(t *testing.T) {
-	checkPeriod = 5 * time.Second
-	wgHandshakeOvertime = 1 * time.Second
-
-	mlog := log.WithField("peer", "tet")
-	mocWgIface := &MocWgIface{}
-	watcher := NewWGWatcher(mlog, mocWgIface, "", newStateDump("peer", mlog, &Status{}))
-
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-
-	ok := watcher.PrepareInitialHandshake()
-	require.True(t, ok, "watcher should not be enabled yet")
-
-	onDisconnected := make(chan struct{}, 1)
-	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
-		mlog.Infof("onDisconnectedFn")
-		onDisconnected <- struct{}{}
-	}, func(when time.Time) {
-		mlog.Infof("onHandshakeSuccess: %v", when)
-	})
-
-	// wait for initial reading
-	time.Sleep(2 * time.Second)
-	mocWgIface.disconnect()
-
-	select {
-	case <-onDisconnected:
-	case <-time.After(10 * time.Second):
-		t.Errorf("timeout")
-	}
-}
-
-func TestWGWatcher_ReEnable(t *testing.T) {
-	checkPeriod = 5 * time.Second
-	wgHandshakeOvertime = 1 * time.Second
-
-	mlog := log.WithField("peer", "tet")
-	mocWgIface := &MocWgIface{}
-	watcher := NewWGWatcher(mlog, mocWgIface, "", newStateDump("peer", mlog, &Status{}))
-
-	ctx, cancel := context.WithCancel(context.Background())
-	ok := watcher.PrepareInitialHandshake()
-	require.True(t, ok, "watcher should not be enabled yet")
-
-	wg := &sync.WaitGroup{}
-	wg.Add(1)
-	go func() {
-		defer wg.Done()
-		watcher.EnableWgWatcher(ctx, time.Now(), func() {}, func(when time.Time) {})
-	}()
-	cancel()
-
-	wg.Wait()
-
-	// Re-enable with a new context
-	ctx, cancel = context.WithCancel(context.Background())
-	defer cancel()
-
-	ok = watcher.PrepareInitialHandshake()
-	require.True(t, ok, "watcher should be re-enabled after the previous run stopped")
-
-	onDisconnected := make(chan struct{}, 1)
-	go watcher.EnableWgWatcher(ctx, time.Now(), func() {
-		onDisconnected <- struct{}{}
-	}, func(when time.Time) {})
-
-	time.Sleep(2 * time.Second)
-	mocWgIface.disconnect()
-
-	select {
-	case <-onDisconnected:
-	case <-time.After(10 * time.Second):
-		t.Errorf("timeout")
-	}
-}

client/internal/peer/worker/priority.go

@@ -1,4 +1,4 @@
-package conntype
+package worker
 
 import (
 	"fmt"

client/internal/peer/worker/worker_ice.go

@@ -1,4 +1,4 @@
-package peer
+package worker
 
 import (
 	"context"
@@ -13,8 +13,9 @@ import (
 
 	"github.com/netbirdio/netbird/client/iface"
 	"github.com/netbirdio/netbird/client/iface/udpmux"
-	"github.com/netbirdio/netbird/client/internal/peer/conntype"
 	icemaker "github.com/netbirdio/netbird/client/internal/peer/ice"
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
+	"github.com/netbirdio/netbird/client/internal/peer/status"
 	"github.com/netbirdio/netbird/client/internal/portforward"
 	"github.com/netbirdio/netbird/client/internal/stdnet"
 	"github.com/netbirdio/netbird/route"
@@ -32,57 +33,68 @@ type ICEConnInfo struct {
 	RelayedOnLocal             bool
 }
 
-type WorkerICE struct {
-	ctx               context.Context
-	log               *log.Entry
-	config            ConnConfig
-	conn              *Conn
-	signaler          *Signaler
-	iFaceDiscover     stdnet.ExternalIFaceDiscover
-	statusRecorder    *Status
-	hasRelayOnLocally bool
+type ICEDependencies struct {
+	Signaler           *signaling.Signaler
+	IFaceDiscover      stdnet.ExternalIFaceDiscover
+	StatusRecorder     *status.Recorder
+	PortForwardManager *portforward.Manager
+}
+
+type ICE struct {
+	log                *log.Entry
+	key                string
+	iceConfig          icemaker.Config
+	isController       bool
+	onConnReady        func(priority ConnPriority, iceConnInfo ICEConnInfo)
+	onStatusDisconnect func(sessionChanged bool)
+	signaler           *signaling.Signaler
+	iFaceDiscover      stdnet.ExternalIFaceDiscover
+	statusRecorder     *status.Recorder
+	portForwardManager *portforward.Manager
+	hasRelayOnLocally  bool
 
 	agent             *icemaker.ThreadSafeAgent
 	agentDialerCancel context.CancelFunc
 	agentConnecting   bool      // while it is true, drop all incoming offers
 	lastSuccess       time.Time // with this avoid the too frequent ICE agent recreation
+	// connectedAgent is the agent whose connection was last reported ready; guarded by muxAgent
+	connectedAgent *icemaker.ThreadSafeAgent
 	// remoteSessionID represents the peer's session identifier from the latest remote offer.
-	remoteSessionID ICESessionID
+	remoteSessionID icemaker.SessionID
 	// sessionID is used to track the current session ID of the ICE agent
 	// increase by one when disconnecting the agent
 	// with it the remote peer can discard the already deprecated offer/answer
 	// Without it the remote peer may recreate a workable ICE connection
-	sessionID            ICESessionID
+	sessionID            icemaker.SessionID
 	remoteSessionChanged bool
 	muxAgent             sync.Mutex
 
 	localUfrag string
 	localPwd   string
 
-	// we record the last known state of the ICE agent to avoid duplicate on disconnected events
-	lastKnownState ice.ConnectionState
-
 	// portForwardAttempted tracks if we've already tried port forwarding this session
 	portForwardAttempted bool
 }
 
-func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *Conn, signaler *Signaler, ifaceDiscover stdnet.ExternalIFaceDiscover, statusRecorder *Status, hasRelayOnLocally bool) (*WorkerICE, error) {
-	sessionID, err := NewICESessionID()
+func NewICE(log *log.Entry, key string, iceConfig icemaker.Config, isController bool, onConnReady func(ConnPriority, ICEConnInfo), onStatusDisconnect func(bool), services ICEDependencies, hasRelayOnLocally bool) (*ICE, error) {
+	sessionID, err := icemaker.NewSessionID()
 	if err != nil {
 		return nil, err
 	}
 
-	w := &WorkerICE{
-		ctx:               ctx,
-		log:               log,
-		config:            config,
-		conn:              conn,
-		signaler:          signaler,
-		iFaceDiscover:     ifaceDiscover,
-		statusRecorder:    statusRecorder,
-		hasRelayOnLocally: hasRelayOnLocally,
-		lastKnownState:    ice.ConnectionStateDisconnected,
-		sessionID:         sessionID,
+	w := &ICE{
+		log:                log,
+		key:                key,
+		iceConfig:          iceConfig,
+		isController:       isController,
+		onConnReady:        onConnReady,
+		onStatusDisconnect: onStatusDisconnect,
+		signaler:           services.Signaler,
+		iFaceDiscover:      services.IFaceDiscover,
+		statusRecorder:     services.StatusRecorder,
+		portForwardManager: services.PortForwardManager,
+		hasRelayOnLocally:  hasRelayOnLocally,
+		sessionID:          sessionID,
 	}
 
 	localUfrag, localPwd, err := icemaker.GenerateICECredentials()
@@ -94,7 +106,7 @@ func NewWorkerICE(ctx context.Context, log *log.Entry, config ConnConfig, conn *
 	return w, nil
 }
 
-func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
+func (w *ICE) OnNewOffer(ctx context.Context, remoteOfferAnswer *signaling.OfferAnswer) {
 	w.log.Debugf("OnNewOffer for ICE, serial: %s", remoteOfferAnswer.SessionIDString())
 	w.muxAgent.Lock()
 	defer w.muxAgent.Unlock()
@@ -118,7 +130,7 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 			}
 		}
 
-		sessionID, err := NewICESessionID()
+		sessionID, err := icemaker.NewSessionID()
 		if err != nil {
 			w.log.Errorf("failed to create new session ID: %s", err)
 		}
@@ -136,8 +148,8 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 	if remoteOfferAnswer.SessionID != nil {
 		w.log.Debugf("recreate ICE agent: %s / %s", w.sessionID, *remoteOfferAnswer.SessionID)
 	}
-	dialerCtx, dialerCancel := context.WithCancel(w.ctx)
-	agent, err := w.reCreateAgent(dialerCancel, preferredCandidateTypes)
+	dialerCtx, dialerCancel := context.WithCancel(ctx)
+	agent, err := w.reCreateAgent(ctx, dialerCancel, preferredCandidateTypes)
 	if err != nil {
 		w.log.Errorf("failed to recreate ICE Agent: %s", err)
 		return
@@ -151,14 +163,14 @@ func (w *WorkerICE) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 		w.remoteSessionID = ""
 	}
 
-	go w.connect(dialerCtx, agent, remoteOfferAnswer)
+	go w.connect(dialerCtx, dialerCancel, agent, remoteOfferAnswer)
 }
 
 // OnRemoteCandidate Handles ICE connection Candidate provided by the remote peer.
-func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
+func (w *ICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HAMap) {
 	w.muxAgent.Lock()
 	defer w.muxAgent.Unlock()
-	w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.config.Key, candidate.String())
+	w.log.Debugf("OnRemoteCandidate from peer %s -> %s", w.key, candidate.String())
 	if w.agent == nil {
 		w.log.Warnf("ICE Agent is not initialized yet")
 		return
@@ -185,18 +197,24 @@ func (w *WorkerICE) OnRemoteCandidate(candidate ice.Candidate, haRoutes route.HA
 	}
 }
 
-func (w *WorkerICE) GetLocalUserCredentials() (frag string, pwd string) {
-	return w.localUfrag, w.localPwd
+func (w *ICE) Credentials() signaling.Credentials {
+	w.muxAgent.Lock()
+	defer w.muxAgent.Unlock()
+	return signaling.Credentials{
+		UFrag:     w.localUfrag,
+		Pwd:       w.localPwd,
+		SessionID: w.sessionID,
+	}
 }
 
-func (w *WorkerICE) InProgress() bool {
+func (w *ICE) InProgress() bool {
 	w.muxAgent.Lock()
 	defer w.muxAgent.Unlock()
 
 	return w.agentConnecting
 }
 
-func (w *WorkerICE) Close() {
+func (w *ICE) Close() {
 	w.muxAgent.Lock()
 	defer w.muxAgent.Unlock()
 
@@ -212,10 +230,10 @@ func (w *WorkerICE) Close() {
 	w.agent = nil
 }
 
-func (w *WorkerICE) reCreateAgent(dialerCancel context.CancelFunc, candidates []ice.CandidateType) (*icemaker.ThreadSafeAgent, error) {
+func (w *ICE) reCreateAgent(ctx context.Context, dialerCancel context.CancelFunc, candidates []ice.CandidateType) (*icemaker.ThreadSafeAgent, error) {
 	w.portForwardAttempted = false
 
-	agent, err := icemaker.NewAgent(w.ctx, w.iFaceDiscover, w.config.ICEConfig, candidates, w.localUfrag, w.localPwd)
+	agent, err := icemaker.NewAgent(ctx, w.iFaceDiscover, w.iceConfig, candidates, w.localUfrag, w.localPwd)
 	if err != nil {
 		return nil, fmt.Errorf("create agent: %w", err)
 	}
@@ -237,7 +255,7 @@ func (w *WorkerICE) reCreateAgent(dialerCancel context.CancelFunc, candidates []
 	return agent, nil
 }
 
-func (w *WorkerICE) SessionID() ICESessionID {
+func (w *ICE) getSessionID() icemaker.SessionID {
 	w.muxAgent.Lock()
 	defer w.muxAgent.Unlock()
 
@@ -247,11 +265,11 @@ func (w *WorkerICE) SessionID() ICESessionID {
 // will block until connection succeeded
 // but it won't release if ICE Agent went into Disconnected or Failed state,
 // so we have to cancel it with the provided context once agent detected a broken connection
-func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) {
+func (w *ICE) connect(ctx context.Context, dialerCancel context.CancelFunc, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *signaling.OfferAnswer) {
 	w.log.Debugf("gather candidates")
 	if err := agent.GatherCandidates(); err != nil {
 		w.log.Warnf("failed to gather candidates: %s", err)
-		w.closeAgent(agent, w.agentDialerCancel)
+		w.closeAgent(agent, dialerCancel)
 		return
 	}
 
@@ -259,19 +277,19 @@ func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent
 	remoteConn, err := w.turnAgentDial(ctx, agent, remoteOfferAnswer)
 	if err != nil {
 		w.log.Debugf("failed to dial the remote peer: %s", err)
-		w.closeAgent(agent, w.agentDialerCancel)
+		w.closeAgent(agent, dialerCancel)
 		return
 	}
 	w.log.Debugf("agent dial succeeded")
 
 	pair, err := agent.GetSelectedCandidatePair()
 	if err != nil {
-		w.closeAgent(agent, w.agentDialerCancel)
+		w.closeAgent(agent, dialerCancel)
 		return
 	}
 	if pair == nil {
 		w.log.Warnf("selected candidate pair is nil, cannot proceed")
-		w.closeAgent(agent, w.agentDialerCancel)
+		w.closeAgent(agent, dialerCancel)
 		return
 	}
 
@@ -299,17 +317,22 @@ func (w *WorkerICE) connect(ctx context.Context, agent *icemaker.ThreadSafeAgent
 	}
 	w.log.Debugf("on ICE conn is ready to use")
 
-	w.log.Infof("connection succeeded with offer session: %s", remoteOfferAnswer.SessionIDString())
 	w.muxAgent.Lock()
+	if w.agent != agent {
+		w.muxAgent.Unlock()
+		w.log.Debugf("agent has been replaced during connect, dropping obsolete connection")
+		return
+	}
 	w.agentConnecting = false
 	w.lastSuccess = time.Now()
+	w.connectedAgent = agent
 	w.muxAgent.Unlock()
 
-	// todo: the potential problem is a race between the onConnectionStateChange
-	w.conn.onICEConnectionIsReady(selectedPriority(pair), ci)
+	w.log.Infof("connection succeeded with offer session: %s", remoteOfferAnswer.SessionIDString())
+	w.onConnReady(selectedPriority(pair), ci)
 }
 
-func (w *WorkerICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.CancelFunc) bool {
+func (w *ICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.CancelFunc) bool {
 	cancel()
 	if err := agent.Close(); err != nil {
 		w.log.Warnf("failed to close ICE agent: %s", err)
@@ -323,7 +346,7 @@ func (w *WorkerICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.C
 
 	if w.agent == agent {
 		// consider to remove from here and move to the OnNewOffer
-		sessionID, err := NewICESessionID()
+		sessionID, err := icemaker.NewSessionID()
 		if err != nil {
 			w.log.Errorf("failed to create new session ID: %s", err)
 		}
@@ -335,7 +358,7 @@ func (w *WorkerICE) closeAgent(agent *icemaker.ThreadSafeAgent, cancel context.C
 	return sessionChanged
 }
 
-func (w *WorkerICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int) {
+func (w *ICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int) {
 	// wait local endpoint configuration
 	time.Sleep(time.Second)
 	addr, err := net.ResolveUDPAddr("udp", net.JoinHostPort(pair.Remote.Address(), strconv.Itoa(remoteWgPort)))
@@ -344,7 +367,7 @@ func (w *WorkerICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int)
 		return
 	}
 
-	mux, ok := w.config.ICEConfig.UDPMuxSrflx.(*udpmux.UniversalUDPMuxDefault)
+	mux, ok := w.iceConfig.UDPMuxSrflx.(*udpmux.UniversalUDPMuxDefault)
 	if !ok {
 		w.log.Warn("invalid udp mux conversion")
 		return
@@ -357,7 +380,7 @@ func (w *WorkerICE) punchRemoteWGPort(pair *ice.CandidatePair, remoteWgPort int)
 
 // onICECandidate is a callback attached to an ICE Agent to receive new local connection candidates
 // and then signals them to the remote peer
-func (w *WorkerICE) onICECandidate(candidate ice.Candidate) {
+func (w *ICE) onICECandidate(candidate ice.Candidate) {
 	// nil means candidate gathering has been ended
 	if candidate == nil {
 		return
@@ -366,9 +389,9 @@ func (w *WorkerICE) onICECandidate(candidate ice.Candidate) {
 	// TODO: reported port is incorrect for CandidateTypeHost, makes understanding ICE use via logs confusing as port is ignored
 	w.log.Debugf("discovered local candidate %s", candidate.String())
 	go func() {
-		err := w.signaler.SignalICECandidate(candidate, w.config.Key)
+		err := w.signaler.SignalICECandidate(candidate, w.key)
 		if err != nil {
-			w.log.Errorf("failed signaling candidate to the remote peer %s %s", w.config.Key, err)
+			w.log.Errorf("failed signaling candidate to the remote peer %s %s", w.key, err)
 		}
 	}()
 
@@ -378,8 +401,8 @@ func (w *WorkerICE) onICECandidate(candidate ice.Candidate) {
 }
 
 // injectPortForwardedCandidate signals an additional candidate using the pre-created port mapping.
-func (w *WorkerICE) injectPortForwardedCandidate(srflxCandidate ice.Candidate) {
-	pfManager := w.conn.portForwardManager
+func (w *ICE) injectPortForwardedCandidate(srflxCandidate ice.Candidate) {
+	pfManager := w.portForwardManager
 	if pfManager == nil {
 		return
 	}
@@ -407,7 +430,7 @@ func (w *WorkerICE) injectPortForwardedCandidate(srflxCandidate ice.Candidate) {
 		forwardedCandidate.String(), mapping.InternalPort, mapping.ExternalPort, mapping.NATType, forwardedCandidate.Priority())
 
 	go func() {
-		if err := w.signaler.SignalICECandidate(forwardedCandidate, w.config.Key); err != nil {
+		if err := w.signaler.SignalICECandidate(forwardedCandidate, w.key); err != nil {
 			w.log.Errorf("signal port-forwarded candidate: %v", err)
 		}
 	}()
@@ -415,7 +438,7 @@ func (w *WorkerICE) injectPortForwardedCandidate(srflxCandidate ice.Candidate) {
 
 // createForwardedCandidate creates a new server reflexive candidate with the forwarded port.
 // It uses the NAT gateway's external IP with the forwarded port.
-func (w *WorkerICE) createForwardedCandidate(srflxCandidate ice.Candidate, mapping *portforward.Mapping) (ice.Candidate, error) {
+func (w *ICE) createForwardedCandidate(srflxCandidate ice.Candidate, mapping *portforward.Mapping) (ice.Candidate, error) {
 	var externalIP string
 	if mapping.ExternalIP != nil && !mapping.ExternalIP.IsUnspecified() {
 		externalIP = mapping.ExternalIP.String()
@@ -460,9 +483,9 @@ func (w *WorkerICE) createForwardedCandidate(srflxCandidate ice.Candidate, mappi
 	return candidate, nil
 }
 
-func (w *WorkerICE) onICESelectedCandidatePair(agent *icemaker.ThreadSafeAgent, c1, c2 ice.Candidate) {
+func (w *ICE) onICESelectedCandidatePair(agent *icemaker.ThreadSafeAgent, c1, c2 ice.Candidate) {
 	w.log.Debugf("selected candidate pair [local <-> remote] -> [%s <-> %s], peer %s", c1.String(), c2.String(),
-		w.config.Key)
+		w.key)
 
 	pairStat, ok := agent.GetSelectedCandidatePairStats()
 	if !ok {
@@ -471,14 +494,14 @@ func (w *WorkerICE) onICESelectedCandidatePair(agent *icemaker.ThreadSafeAgent,
 	}
 
 	duration := time.Duration(pairStat.CurrentRoundTripTime * float64(time.Second))
-	if err := w.statusRecorder.UpdateLatency(w.config.Key, duration); err != nil {
+	if err := w.statusRecorder.UpdateLatency(w.key, duration); err != nil {
 		w.log.Debugf("failed to update latency for peer: %s", err)
 		return
 	}
 }
 
-func (w *WorkerICE) logSuccessfulPaths(agent *icemaker.ThreadSafeAgent) {
-	sessionID := w.SessionID()
+func (w *ICE) logSuccessfulPaths(agent *icemaker.ThreadSafeAgent) {
+	sessionID := w.getSessionID()
 	stats := agent.GetCandidatePairsStats()
 	localCandidates, _ := agent.GetLocalCandidates()
 	remoteCandidates, _ := agent.GetRemoteCandidates()
@@ -508,32 +531,44 @@ func (w *WorkerICE) logSuccessfulPaths(agent *icemaker.ThreadSafeAgent) {
 	}
 }
 
-func (w *WorkerICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dialerCancel context.CancelFunc) func(ice.ConnectionState) {
+func (w *ICE) onConnectionStateChange(agent *icemaker.ThreadSafeAgent, dialerCancel context.CancelFunc) func(ice.ConnectionState) {
+	// per-agent state; pion delivers callbacks of one agent sequentially
+	var connected bool
 	return func(state ice.ConnectionState) {
 		w.log.Debugf("ICE ConnectionState has changed to %s", state.String())
 		switch state {
 		case ice.ConnectionStateConnected:
-			w.lastKnownState = ice.ConnectionStateConnected
+			connected = true
 			w.logSuccessfulPaths(agent)
-			return
 		case ice.ConnectionStateFailed, ice.ConnectionStateDisconnected, ice.ConnectionStateClosed:
 			// ice.ConnectionStateClosed happens when we recreate the agent. For the P2P to TURN switch important to
 			// notify the conn.onICEStateDisconnected changes to update the current used priority
 
 			sessionChanged := w.closeAgent(agent, dialerCancel)
 
-			if w.lastKnownState == ice.ConnectionStateConnected {
-				w.lastKnownState = ice.ConnectionStateDisconnected
-				w.conn.onICEStateDisconnected(sessionChanged)
+			if !connected {
+				return
 			}
-		default:
-			return
+			connected = false
+
+			w.muxAgent.Lock()
+			stale := w.connectedAgent != agent
+			if !stale {
+				w.connectedAgent = nil
+			}
+			w.muxAgent.Unlock()
+
+			if stale {
+				w.log.Debugf("suppress disconnected event of replaced ICE agent")
+				return
+			}
+			w.onStatusDisconnect(sessionChanged)
 		}
 	}
 }
 
-func (w *WorkerICE) turnAgentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *OfferAnswer) (*ice.Conn, error) {
-	if isController(w.config) {
+func (w *ICE) turnAgentDial(ctx context.Context, agent *icemaker.ThreadSafeAgent, remoteOfferAnswer *signaling.OfferAnswer) (*ice.Conn, error) {
+	if w.isController {
 		return agent.Dial(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
 	} else {
 		return agent.Accept(ctx, remoteOfferAnswer.IceCredentials.UFrag, remoteOfferAnswer.IceCredentials.Pwd)
@@ -595,10 +630,10 @@ func isRelayed(pair *ice.CandidatePair) bool {
 	return false
 }
 
-func selectedPriority(pair *ice.CandidatePair) conntype.ConnPriority {
+func selectedPriority(pair *ice.CandidatePair) ConnPriority {
 	if isRelayed(pair) {
-		return conntype.ICETurn
+		return ICETurn
 	} else {
-		return conntype.ICEP2P
+		return ICEP2P
 	}
 }

client/internal/peer/worker/worker_relay.go

@@ -1,4 +1,4 @@
-package peer
+package worker
 
 import (
 	"context"
@@ -10,22 +10,23 @@ import (
 
 	log "github.com/sirupsen/logrus"
 
+	"github.com/netbirdio/netbird/client/internal/peer/signaling"
 	relayClient "github.com/netbirdio/netbird/shared/relay/client"
 )
 
 type RelayConnInfo struct {
-	relayedConn     net.Conn
-	rosenpassPubKey []byte
-	rosenpassAddr   string
+	RelayedConn     net.Conn
+	RosenpassPubKey []byte
+	RosenpassAddr   string
 }
 
 type WorkerRelay struct {
-	peerCtx      context.Context
-	log          *log.Entry
-	isController bool
-	config       ConnConfig
-	conn         *Conn
-	relayManager *relayClient.Manager
+	log            *log.Entry
+	key            string
+	isController   bool
+	onConnReady    func(RelayConnInfo)
+	onDisconnected func()
+	relayManager   *relayClient.Manager
 
 	relayedConn net.Conn
 	relayLock   sync.Mutex
@@ -33,19 +34,19 @@ type WorkerRelay struct {
 	relaySupportedOnRemotePeer atomic.Bool
 }
 
-func NewWorkerRelay(ctx context.Context, log *log.Entry, ctrl bool, config ConnConfig, conn *Conn, relayManager *relayClient.Manager) *WorkerRelay {
+func NewWorkerRelay(log *log.Entry, key string, isController bool, onConnReady func(RelayConnInfo), onDisconnected func(), relayManager *relayClient.Manager) *WorkerRelay {
 	r := &WorkerRelay{
-		peerCtx:      ctx,
-		log:          log,
-		isController: ctrl,
-		config:       config,
-		conn:         conn,
-		relayManager: relayManager,
+		log:            log,
+		key:            key,
+		isController:   isController,
+		onConnReady:    onConnReady,
+		onDisconnected: onDisconnected,
+		relayManager:   relayManager,
 	}
 	return r
 }
 
-func (w *WorkerRelay) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
+func (w *WorkerRelay) OnNewOffer(ctx context.Context, remoteOfferAnswer *signaling.OfferAnswer) {
 	if !w.isRelaySupported(remoteOfferAnswer) {
 		w.log.Infof("Relay is not supported by remote peer")
 		w.relaySupportedOnRemotePeer.Store(false)
@@ -66,7 +67,7 @@ func (w *WorkerRelay) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 		serverIP = remoteOfferAnswer.RelaySrvIP
 	}
 
-	relayedConn, err := w.relayManager.OpenConn(w.peerCtx, srv, w.config.Key, serverIP)
+	relayedConn, err := w.relayManager.OpenConn(ctx, srv, w.key, serverIP)
 	if err != nil {
 		if errors.Is(err, relayClient.ErrConnAlreadyExists) {
 			w.log.Debugf("handled offer by reusing existing relay connection")
@@ -88,10 +89,10 @@ func (w *WorkerRelay) OnNewOffer(remoteOfferAnswer *OfferAnswer) {
 	}
 
 	w.log.Debugf("peer conn opened via Relay: %s", srv)
-	go w.conn.onRelayConnectionIsReady(RelayConnInfo{
-		relayedConn:     relayedConn,
-		rosenpassPubKey: remoteOfferAnswer.RosenpassPubKey,
-		rosenpassAddr:   remoteOfferAnswer.RosenpassAddr,
+	w.onConnReady(RelayConnInfo{
+		RelayedConn:     relayedConn,
+		RosenpassPubKey: remoteOfferAnswer.RosenpassPubKey,
+		RosenpassAddr:   remoteOfferAnswer.RosenpassAddr,
 	})
 }
 
@@ -119,7 +120,7 @@ func (w *WorkerRelay) CloseConn() {
 	}
 }
 
-func (w *WorkerRelay) isRelaySupported(answer *OfferAnswer) bool {
+func (w *WorkerRelay) isRelaySupported(answer *signaling.OfferAnswer) bool {
 	if !w.relayManager.HasRelayAddress() {
 		return false
 	}
@@ -134,5 +135,5 @@ func (w *WorkerRelay) preferredRelayServer(myRelayAddress, remoteRelayAddress st
 }
 
 func (w *WorkerRelay) onRelayClientDisconnected() {
-	go w.conn.onRelayDisconnected()
+	w.onDisconnected()
 }

client/internal/peer/worker_status.go

@@ -1,4 +1,4 @@
-package worker
+package peer
 
 import (
 	"sync/atomic"
@@ -7,17 +7,17 @@ import (
 )
 
 const (
-	StatusDisconnected Status = iota
-	StatusConnected
+	WorkerStatusDisconnected WorkerStatus = iota
+	WorkerStatusConnected
 )
 
-type Status int32
+type WorkerStatus int32
 
-func (s Status) String() string {
+func (s WorkerStatus) String() string {
 	switch s {
-	case StatusDisconnected:
+	case WorkerStatusDisconnected:
 		return "Disconnected"
-	case StatusConnected:
+	case WorkerStatusConnected:
 		return "Connected"
 	default:
 		log.Errorf("unknown status: %d", s)
@@ -37,16 +37,16 @@ func NewAtomicStatus() *AtomicWorkerStatus {
 }
 
 // Get returns the current connection status
-func (acs *AtomicWorkerStatus) Get() Status {
-	return Status(acs.status.Load())
+func (acs *AtomicWorkerStatus) Get() WorkerStatus {
+	return WorkerStatus(acs.status.Load())
 }
 
 func (acs *AtomicWorkerStatus) SetConnected() {
-	acs.status.Store(int32(StatusConnected))
+	acs.status.Store(int32(WorkerStatusConnected))
 }
 
 func (acs *AtomicWorkerStatus) SetDisconnected() {
-	acs.status.Store(int32(StatusDisconnected))
+	acs.status.Store(int32(WorkerStatusDisconnected))
 }
 
 // String returns the string representation of the current status

client/internal/profilemanager/service.go

@@ -11,6 +11,7 @@ import (
 	"runtime"
 	"sort"
 	"strings"
+	"syscall"
 
 	log "github.com/sirupsen/logrus"
 
@@ -439,7 +440,11 @@ func (s *ServiceManager) GetStatePath() string {
 
 	activeProf, err := s.GetActiveProfileState()
 	if err != nil {
-		log.Warnf("failed to get active profile state: %v", err)
+		if errors.Is(err, syscall.ENOSYS) {
+			log.Debugf("active profile state unavailable on this platform: %v", err)
+		} else {
+			log.Warnf("failed to get active profile state: %v", err)
+		}
 		return defaultStatePath
 	}
 

client/internal/profilemanager/state.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"os"
 	"path/filepath"
 
 	"github.com/netbirdio/netbird/util"
@@ -71,3 +72,22 @@ func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
 
 	return nil
 }
+
+// RemoveProfileState deletes the per-profile state file (which holds the
+// account email used for the SSO login hint and the UI display). Called after
+// a successful logout so a logged-out profile no longer shows a stale account
+// email. The state file only stores the email, so deleting it is equivalent to
+// clearing it; the next SSO login recreates it. A missing file is not an error.
+func (pm *ProfileManager) RemoveProfileState(profileName string) error {
+	configDir, err := getConfigDir()
+	if err != nil {
+		return fmt.Errorf("get config directory: %w", err)
+	}
+
+	stateFile := filepath.Join(configDir, profileName+".state.json")
+	if err := os.Remove(stateFile); err != nil && !os.IsNotExist(err) {
+		return fmt.Errorf("remove profile state: %w", err)
+	}
+
+	return nil
+}

client/internal/rosenpass/manager.go

@@ -39,6 +39,7 @@ type rpServer interface {
 
 type Manager struct {
 	ifaceName    string
+	localWgKey   wgtypes.Key
 	spk          []byte
 	ssk          []byte
 	rpKeyHash    string
@@ -51,8 +52,9 @@ type Manager struct {
 	wgIface      PresharedKeySetter
 }
 
-// NewManager creates a new Rosenpass manager
-func NewManager(preSharedKey *wgtypes.Key, wgIfaceName string) (*Manager, error) {
+// NewManager creates a new Rosenpass manager. localWgKey is the local
+// WireGuard public key, used to derive the per-peer rendezvous key.
+func NewManager(preSharedKey *wgtypes.Key, wgIfaceName string, localWgKey wgtypes.Key) (*Manager, error) {
 	public, secret, err := rp.GenerateKeyPair()
 	if err != nil {
 		return nil, err
@@ -62,6 +64,7 @@ func NewManager(preSharedKey *wgtypes.Key, wgIfaceName string) (*Manager, error)
 	log.Tracef("generated new rosenpass key pair with public key %s", rpKeyHash)
 	return &Manager{
 		ifaceName:    wgIfaceName,
+		localWgKey:   localWgKey,
 		rpKeyHash:    rpKeyHash,
 		spk:          public,
 		ssk:          secret,
@@ -73,7 +76,7 @@ func NewManager(preSharedKey *wgtypes.Key, wgIfaceName string) (*Manager, error)
 		// nil receiver in addPeer -> m.rpWgHandler.AddPeer. generateConfig will
 		// replace it with a fresh handler on each Run() to clear stale peer
 		// state from previous engine sessions.
-		rpWgHandler: NewNetbirdHandler(),
+		rpWgHandler: NewNetbirdHandler((*[32]byte)(preSharedKey), localWgKey),
 		lock:        sync.Mutex{},
 	}, nil
 }
@@ -161,7 +164,7 @@ func (m *Manager) generateConfig() (rp.Config, error) {
 	cfg.Peers = []rp.PeerConfig{}
 
 	m.lock.Lock()
-	m.rpWgHandler = NewNetbirdHandler()
+	m.rpWgHandler = NewNetbirdHandler(m.preSharedKey, m.localWgKey)
 	if m.wgIface != nil {
 		m.rpWgHandler.SetInterface(m.wgIface)
 	}

client/internal/rosenpass/manager_test.go

@@ -85,7 +85,7 @@ func newTestManager(spkFirstByte byte, mock *mockServer) *Manager {
 		ssk:         make([]byte, 32),
 		rpKeyHash:   "test-hash",
 		rpPeerIDs:   make(map[string]*rp.PeerID),
-		rpWgHandler: NewNetbirdHandler(),
+		rpWgHandler: NewNetbirdHandler(nil, wgtypes.Key{0x01}),
 		server:      mock,
 	}
 }
@@ -255,7 +255,7 @@ func TestAddPeer_NilServer_ReturnsErrorNoCrash(t *testing.T) {
 // issue #4341 cannot occur in the window between NewManager and Run().
 func TestNewManager_PreInitializesHandler(t *testing.T) {
 	psk := wgtypes.Key{}
-	m, err := NewManager(&psk, "wt0")
+	m, err := NewManager(&psk, "wt0", wgtypes.Key{0x01})
 	require.NoError(t, err)
 	require.NotNil(t, m.rpWgHandler, "rpWgHandler must be initialized in NewManager")
 }
@@ -329,10 +329,10 @@ func TestIsPresharedKeyInitialized_AddedButNotHandshaken_ReturnsFalse(t *testing
 	require.False(t, m.IsPresharedKeyInitialized(wgKey))
 }
 
-// --- NetbirdHandler.outputKey ----------------------------------------------
+// --- NetbirdHandler.applyKey ----------------------------------------------
 
-func TestHandler_OutputKey_FirstCallUsesUpdateOnlyFalse(t *testing.T) {
-	h := NewNetbirdHandler()
+func TestHandler_ApplyKey_FirstCallUsesUpdateOnlyFalse(t *testing.T) {
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	iface := &mockIface{}
 	h.SetInterface(iface)
 
@@ -348,8 +348,8 @@ func TestHandler_OutputKey_FirstCallUsesUpdateOnlyFalse(t *testing.T) {
 	require.Equal(t, wgKey.String(), iface.calls[0].peerKey)
 }
 
-func TestHandler_OutputKey_SubsequentCallsUseUpdateOnlyTrue(t *testing.T) {
-	h := NewNetbirdHandler()
+func TestHandler_ApplyKey_SubsequentCallsUseUpdateOnlyTrue(t *testing.T) {
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	iface := &mockIface{}
 	h.SetInterface(iface)
 
@@ -364,8 +364,8 @@ func TestHandler_OutputKey_SubsequentCallsUseUpdateOnlyTrue(t *testing.T) {
 	require.True(t, iface.calls[1].updateOnly, "subsequent rotations must use updateOnly=true")
 }
 
-func TestHandler_OutputKey_NilInterface_NoCrashNoCall(t *testing.T) {
-	h := NewNetbirdHandler()
+func TestHandler_ApplyKey_NilInterface_NoCrashNoCall(t *testing.T) {
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	// no SetInterface — iface remains nil
 	pid := rp.PeerID{0x03}
 	h.AddPeer(pid, "wt0", rp.Key(wgtypes.Key{}))
@@ -374,8 +374,8 @@ func TestHandler_OutputKey_NilInterface_NoCrashNoCall(t *testing.T) {
 	h.HandshakeCompleted(pid, rp.Key{})
 }
 
-func TestHandler_OutputKey_UnknownPeer_NoCall(t *testing.T) {
-	h := NewNetbirdHandler()
+func TestHandler_ApplyKey_UnknownPeer_NoCall(t *testing.T) {
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	iface := &mockIface{}
 	h.SetInterface(iface)
 
@@ -384,7 +384,7 @@ func TestHandler_OutputKey_UnknownPeer_NoCall(t *testing.T) {
 }
 
 func TestHandler_RemovePeer_ClearsInitializedState(t *testing.T) {
-	h := NewNetbirdHandler()
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	iface := &mockIface{}
 	h.SetInterface(iface)
 
@@ -398,7 +398,7 @@ func TestHandler_RemovePeer_ClearsInitializedState(t *testing.T) {
 }
 
 func TestHandler_SetInterfaceAfterAddPeer_StillReceivesKey(t *testing.T) {
-	h := NewNetbirdHandler()
+	h := NewNetbirdHandler(nil, wgtypes.Key{0x01})
 	pid := rp.PeerID{0x05}
 	wgKey := wgtypes.Key{0xEE}
 	h.AddPeer(pid, "wt0", rp.Key(wgKey))

client/internal/rosenpass/netbird_handler.go

@@ -18,19 +18,34 @@ type PresharedKeySetter interface {
 type wireGuardPeer struct {
 	Interface string
 	PublicKey rp.Key
+	// initialized is true once a completed exchange has set a
+	// Rosenpass-managed PSK for this peer.
+	initialized bool
+	// chainKey is the key output by the last completed exchange, advanced by
+	// one ratchet step on expiry. Nil until the first exchange completes and
+	// after the peer has fallen back to the rendezvous key.
+	chainKey *wgtypes.Key
+	// expiries counts failed renewals since the last completed exchange.
+	expiries int
 }
 
 type NetbirdHandler struct {
-	mu               sync.Mutex
-	iface            PresharedKeySetter
-	peers            map[rp.PeerID]wireGuardPeer
-	initializedPeers map[rp.PeerID]bool
+	mu    sync.Mutex
+	iface PresharedKeySetter
+	// preSharedKey is the account-level preshared key, used as the rendezvous
+	// key when set. Nil means the deterministic seed key is used instead.
+	preSharedKey *[32]byte
+	// localWgKey is the local WireGuard public key, one of the two inputs to
+	// the deterministic seed key.
+	localWgKey wgtypes.Key
+	peers      map[rp.PeerID]*wireGuardPeer
 }
 
-func NewNetbirdHandler() *NetbirdHandler {
+func NewNetbirdHandler(preSharedKey *[32]byte, localWgKey wgtypes.Key) *NetbirdHandler {
 	return &NetbirdHandler{
-		peers:            map[rp.PeerID]wireGuardPeer{},
-		initializedPeers: map[rp.PeerID]bool{},
+		preSharedKey: preSharedKey,
+		localWgKey:   localWgKey,
+		peers:        map[rp.PeerID]*wireGuardPeer{},
 	}
 }
 
@@ -42,10 +57,16 @@ func (h *NetbirdHandler) SetInterface(iface PresharedKeySetter) {
 	h.iface = iface
 }
 
+// AddPeer registers a peer with the handler. Re-adding a known peer (every
+// reconnection does) keeps its key recovery state.
 func (h *NetbirdHandler) AddPeer(pid rp.PeerID, intf string, pk rp.Key) {
 	h.mu.Lock()
 	defer h.mu.Unlock()
-	h.peers[pid] = wireGuardPeer{
+	if existing, ok := h.peers[pid]; ok && existing.PublicKey == pk {
+		existing.Interface = intf
+		return
+	}
+	h.peers[pid] = &wireGuardPeer{
 		Interface: intf,
 		PublicKey: pk,
 	}
@@ -55,7 +76,6 @@ func (h *NetbirdHandler) RemovePeer(pid rp.PeerID) {
 	h.mu.Lock()
 	defer h.mu.Unlock()
 	delete(h.peers, pid)
-	delete(h.initializedPeers, pid)
 }
 
 // IsPeerInitialized returns true if Rosenpass has completed a handshake
@@ -63,50 +83,120 @@ func (h *NetbirdHandler) RemovePeer(pid rp.PeerID) {
 func (h *NetbirdHandler) IsPeerInitialized(pid rp.PeerID) bool {
 	h.mu.Lock()
 	defer h.mu.Unlock()
-	return h.initializedPeers[pid]
+	peer, ok := h.peers[pid]
+	return ok && peer.initialized
 }
 
+// HandshakeCompleted programs the freshly exchanged output key and resets the
+// peer's key recovery state.
 func (h *NetbirdHandler) HandshakeCompleted(pid rp.PeerID, key rp.Key) {
-	h.outputKey(rp.KeyOutputReasonStale, pid, key)
-}
+	psk := wgtypes.Key(key)
 
-func (h *NetbirdHandler) HandshakeExpired(pid rp.PeerID) {
-	key, _ := rp.GeneratePresharedKey()
-	h.outputKey(rp.KeyOutputReasonStale, pid, key)
-}
-
-func (h *NetbirdHandler) outputKey(_ rp.KeyOutputReason, pid rp.PeerID, psk rp.Key) {
 	h.mu.Lock()
-	iface := h.iface
-	wg, ok := h.peers[pid]
-	isInitialized := h.initializedPeers[pid]
-	h.mu.Unlock()
+	defer h.mu.Unlock()
 
-	if iface == nil {
-		log.Warn("rosenpass: interface not set, cannot update preshared key")
+	peer, ok := h.peers[pid]
+	if !ok {
 		return
 	}
+	if peer.expiries > 0 {
+		log.Infof("rosenpass exchange completed for peer %s after %d expired renewals", wgtypes.Key(peer.PublicKey), peer.expiries)
+	}
+	// chainKey tracks the shared exchange output regardless of the local write
+	// outcome, so both ends still converge on the next expiry.
+	peer.chainKey = &psk
+	peer.expiries = 0
+	if !h.applyKeyLocked(pid, psk, peer.initialized) {
+		return
+	}
+	peer.initialized = true
+}
 
+// HandshakeExpired replaces the expired key. The renewal exchange runs over
+// the tunnel keyed by the PSK itself, so the replacement must be derivable on
+// both ends without communication: the first expiry ratchets the last shared
+// key forward, repeated expiries (and expiries without a completed exchange)
+// fall back to the rendezvous key and drop the peer out of the initialized
+// state so connection reconfigurations reprogram the rendezvous key as well.
+func (h *NetbirdHandler) HandshakeExpired(pid rp.PeerID) {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	peer, ok := h.peers[pid]
 	if !ok {
 		return
 	}
 
-	peerKey := wgtypes.Key(wg.PublicKey).String()
-	pskKey := wgtypes.Key(psk)
+	peer.expiries++
 
-	// Use updateOnly=true for later rotations (peer already has Rosenpass PSK)
-	// Use updateOnly=false for first rotation (peer has original/empty PSK)
-	if err := iface.SetPresharedKey(peerKey, pskKey, isInitialized); err != nil {
+	var psk wgtypes.Key
+	if peer.chainKey != nil && peer.expiries == 1 {
+		log.Infof("rosenpass key for peer %s expired without renewal, advancing to ratcheted key", wgtypes.Key(peer.PublicKey))
+		psk = RatchetKey(*peer.chainKey)
+		peer.chainKey = &psk
+	} else {
+		rendezvous, err := h.rendezvousKey(peer)
+		if err != nil {
+			// Fail closed: without a rendezvous key the expired key must
+			// still be rotated out, even if the replacement is unusable.
+			log.Errorf("failed to derive rendezvous key, replacing expired key with a random one: %v", err)
+			h.applyRandomKeyLocked(pid)
+			return
+		}
+		log.Warnf("rosenpass key for peer %s expired %d times without renewal, falling back to the rendezvous key", wgtypes.Key(peer.PublicKey), peer.expiries)
+		psk = rendezvous
+		peer.chainKey = nil
+		peer.initialized = false
+	}
+
+	h.applyKeyLocked(pid, psk, true)
+}
+
+// rendezvousKey returns the key both ends converge on without communication:
+// the account-level preshared key when configured, the deterministic seed key
+// otherwise. It mirrors the key that peer connections program when Rosenpass
+// does not manage the peer yet.
+func (h *NetbirdHandler) rendezvousKey(peer *wireGuardPeer) (wgtypes.Key, error) {
+	if h.preSharedKey != nil {
+		return *h.preSharedKey, nil
+	}
+
+	seed, err := DeterministicSeedKey(h.localWgKey.String(), wgtypes.Key(peer.PublicKey).String())
+	if err != nil {
+		return wgtypes.Key{}, err
+	}
+	return *seed, nil
+}
+
+// applyKeyLocked writes the preshared key for the peer to the WireGuard
+// interface and reports whether the write succeeded. Callers must hold h.mu
+// for the whole state-mutation-plus-write so that a concurrent completion and
+// expiry cannot reorder their writes relative to the in-memory chain key.
+func (h *NetbirdHandler) applyKeyLocked(pid rp.PeerID, psk wgtypes.Key, updateOnly bool) bool {
+	peer, ok := h.peers[pid]
+	if !ok {
+		return false
+	}
+
+	if h.iface == nil {
+		log.Warn("rosenpass: interface not set, cannot update preshared key")
+		return false
+	}
+
+	peerKey := wgtypes.Key(peer.PublicKey).String()
+	if err := h.iface.SetPresharedKey(peerKey, psk, updateOnly); err != nil {
 		log.Errorf("Failed to apply rosenpass key: %v", err)
+		return false
+	}
+
+	return true
+}
+
+func (h *NetbirdHandler) applyRandomKeyLocked(pid rp.PeerID) {
+	key, err := rp.GeneratePresharedKey()
+	if err != nil {
+		log.Errorf("failed to generate random preshared key: %v", err)
 		return
 	}
-
-	// Mark peer as isInitialized after the successful first rotation
-	if !isInitialized {
-		h.mu.Lock()
-		if _, exists := h.peers[pid]; exists {
-			h.initializedPeers[pid] = true
-		}
-		h.mu.Unlock()
-	}
+	h.applyKeyLocked(pid, wgtypes.Key(key), true)
 }

client/internal/rosenpass/netbird_handler_test.go

@@ -0,0 +1,250 @@
+package rosenpass
+
+import (
+	"testing"
+
+	rp "cunicu.li/go-rosenpass"
+	"github.com/stretchr/testify/require"
+	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
+)
+
+// handlerTestLink wires two NetbirdHandlers as the two ends of a single
+// tunnel: handler A manages the rosenpass peer B and vice versa, the way two
+// NetBird clients see each other.
+type handlerTestLink struct {
+	handlerA, handlerB *NetbirdHandler
+	ifaceA, ifaceB     *mockIface
+	pidA, pidB         rp.PeerID
+	wgKeyA, wgKeyB     wgtypes.Key
+}
+
+func newHandlerTestLink(t *testing.T, preSharedKey *[32]byte) *handlerTestLink {
+	t.Helper()
+
+	link := &handlerTestLink{
+		ifaceA: &mockIface{},
+		ifaceB: &mockIface{},
+	}
+	link.pidA[0] = 0xaa
+	link.pidB[0] = 0xbb
+	link.wgKeyA[31] = 1
+	link.wgKeyB[31] = 2
+
+	link.handlerA = NewNetbirdHandler(preSharedKey, link.wgKeyA)
+	link.handlerB = NewNetbirdHandler(preSharedKey, link.wgKeyB)
+
+	link.handlerA.SetInterface(link.ifaceA)
+	link.handlerB.SetInterface(link.ifaceB)
+
+	link.handlerA.AddPeer(link.pidB, "wt0", rp.Key(link.wgKeyB))
+	link.handlerB.AddPeer(link.pidA, "wt0", rp.Key(link.wgKeyA))
+
+	return link
+}
+
+// complete simulates a completed rosenpass exchange: both ends derive the
+// same output key.
+func (l *handlerTestLink) complete(osk rp.Key) {
+	l.handlerA.HandshakeCompleted(l.pidB, osk)
+	l.handlerB.HandshakeCompleted(l.pidA, osk)
+}
+
+// expire simulates a failed key renewal on both ends.
+func (l *handlerTestLink) expire() {
+	l.handlerA.HandshakeExpired(l.pidB)
+	l.handlerB.HandshakeExpired(l.pidA)
+}
+
+func lastPSK(t *testing.T, m *mockIface) wgtypes.Key {
+	t.Helper()
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	require.NotEmpty(t, m.calls, "expected at least one SetPresharedKey call")
+	return m.calls[len(m.calls)-1].psk
+}
+
+func TestHandshakeCompleted_SetsKeyAndInitializes(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	require.Equal(t, wgtypes.Key(osk), lastPSK(t, link.ifaceA), "completed exchange must program the osk")
+	require.False(t, link.ifaceA.calls[0].updateOnly, "first rotation must not be update-only")
+	require.True(t, link.handlerA.IsPeerInitialized(link.pidB), "peer must be initialized after first completed exchange")
+
+	link.complete(osk)
+	require.True(t, link.ifaceA.calls[1].updateOnly, "later rotations must be update-only")
+}
+
+// TestHandshakeExpired_BothSidesConverge encodes the core recovery invariant:
+// rosenpass renewals run over the tunnel that the PSK itself keys, so when a
+// renewal fails on both ends, both ends must fall back to the same key or the
+// tunnel can never handshake again.
+func TestHandshakeExpired_BothSidesConverge(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	keyA := lastPSK(t, link.ifaceA)
+	keyB := lastPSK(t, link.ifaceB)
+	require.NotEqual(t, wgtypes.Key(osk), keyA, "expired key must be rotated out")
+	require.Equal(t, keyA, keyB, "both ends must converge on the same key after expiry")
+
+	link.expire()
+	require.Equal(t, lastPSK(t, link.ifaceA), lastPSK(t, link.ifaceB),
+		"both ends must still converge after repeated expiries")
+}
+
+// TestHandshakeExpired_ExpiryWithoutCompletionConverges covers the bootstrap
+// case: the initial exchange never completed (the tunnel ran on the rendezvous
+// key), so an expiry must not replace the working key with an unrecoverable
+// one on either end.
+func TestHandshakeExpired_ExpiryWithoutCompletionConverges(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	link.expire()
+	require.Equal(t, lastPSK(t, link.ifaceA), lastPSK(t, link.ifaceB),
+		"both ends must converge when the exchange never completed")
+}
+
+// TestHandshakeExpired_RepeatedExpiryClearsInitialized: once renewals keep
+// failing, the peer must drop out of the initialized state so the next
+// connection reconfiguration reprograms the rendezvous key instead of
+// preserving a poisoned rosenpass-managed key.
+func TestHandshakeExpired_RepeatedExpiryClearsInitialized(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	link.expire()
+
+	require.False(t, link.handlerA.IsPeerInitialized(link.pidB),
+		"repeated expiries must clear the initialized state")
+	require.False(t, link.handlerB.IsPeerInitialized(link.pidA),
+		"repeated expiries must clear the initialized state")
+}
+
+// TestHandshakeCompleted_AfterExpiryRecovers: a completed exchange after a
+// desync must fully reset the recovery state.
+func TestHandshakeCompleted_AfterExpiryRecovers(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk1, osk2 rp.Key
+	osk1[0] = 1
+	osk2[0] = 2
+
+	link.complete(osk1)
+	link.expire()
+	link.expire()
+
+	link.complete(osk2)
+	require.Equal(t, wgtypes.Key(osk2), lastPSK(t, link.ifaceA), "new exchange must program the fresh osk")
+	require.True(t, link.handlerA.IsPeerInitialized(link.pidB), "peer must be initialized again after recovery")
+
+	link.expire()
+	require.Equal(t, lastPSK(t, link.ifaceA), lastPSK(t, link.ifaceB),
+		"recovered link must converge again on the next expiry")
+	require.NotEqual(t, wgtypes.Key(osk2), lastPSK(t, link.ifaceA), "expired key must be rotated out")
+}
+
+// TestHandshakeExpired_FirstExpiryRatchetsLastKey: the first expiry must
+// derive the replacement from the last shared key, so an attacker who only
+// blocks the renewal exchange gains nothing over the previous key.
+func TestHandshakeExpired_FirstExpiryRatchetsLastKey(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	require.Equal(t, RatchetKey(wgtypes.Key(osk)), lastPSK(t, link.ifaceA),
+		"first expiry must program the ratcheted key")
+	require.True(t, link.handlerA.IsPeerInitialized(link.pidB),
+		"ratchet step must keep the peer initialized so reconfigurations preserve the key")
+}
+
+// TestHandshakeExpired_RepeatedExpiryFallsBackToSeed: once the ratchet key
+// also fails, both ends must land on the same key that peer connections
+// program for uninitialized peers, so a reconnect completes the recovery.
+func TestHandshakeExpired_RepeatedExpiryFallsBackToSeed(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	link.expire()
+
+	seed, err := DeterministicSeedKey(link.wgKeyA.String(), link.wgKeyB.String())
+	require.NoError(t, err)
+	require.Equal(t, *seed, lastPSK(t, link.ifaceA), "repeated expiry must fall back to the seed key")
+	require.Equal(t, *seed, lastPSK(t, link.ifaceB), "repeated expiry must fall back to the seed key")
+}
+
+// TestHandshakeExpired_ConfiguredPSKUsedAsRendezvous: with an account-level
+// preshared key configured, the fallback must be that key, matching what peer
+// connections program for uninitialized peers.
+func TestHandshakeExpired_ConfiguredPSKUsedAsRendezvous(t *testing.T) {
+	psk := &[32]byte{0x77}
+	link := newHandlerTestLink(t, psk)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	link.expire()
+
+	require.Equal(t, wgtypes.Key(*psk), lastPSK(t, link.ifaceA),
+		"fallback must be the configured preshared key")
+	require.Equal(t, wgtypes.Key(*psk), lastPSK(t, link.ifaceB),
+		"fallback must be the configured preshared key on both ends")
+}
+
+// TestHandshakeExpired_ExpiryWritesAreUpdateOnly: expiry replacements must
+// never create a WireGuard peer that connection management has removed.
+func TestHandshakeExpired_ExpiryWritesAreUpdateOnly(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+
+	link.expire()
+	link.expire()
+
+	for _, call := range link.ifaceA.calls[1:] {
+		require.True(t, call.updateOnly, "expiry writes must be update-only")
+	}
+}
+
+// TestAddPeer_ReAddKeepsRecoveryState: reconnections re-add the peer on every
+// OnConnected; that must not reset the expiry chain state.
+func TestAddPeer_ReAddKeepsRecoveryState(t *testing.T) {
+	link := newHandlerTestLink(t, nil)
+
+	var osk rp.Key
+	osk[0] = 0x42
+	link.complete(osk)
+	link.expire()
+
+	link.handlerA.AddPeer(link.pidB, "wt0", rp.Key(link.wgKeyB))
+	require.True(t, link.handlerA.IsPeerInitialized(link.pidB),
+		"re-adding a known peer must keep its state")
+
+	link.expire()
+	seed, err := DeterministicSeedKey(link.wgKeyA.String(), link.wgKeyB.String())
+	require.NoError(t, err)
+	require.Equal(t, *seed, lastPSK(t, link.ifaceA),
+		"second expiry after re-add must continue to the seed fallback")
+}

client/internal/rosenpass/seed.go

@@ -1,11 +1,28 @@
 package rosenpass
 
 import (
+	"crypto/sha256"
 	"fmt"
 
 	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
 )
 
+// ratchetLabel domain-separates the expiry ratchet from other uses of the
+// rosenpass output key.
+const ratchetLabel = "netbird-rosenpass-expiry-ratchet"
+
+// RatchetKey derives the successor preshared key from the previous Rosenpass
+// output key. When a key expires without a completed renewal, both peers
+// advance their last shared key by one ratchet step: the expired key is
+// rotated out while both ends still converge on an identical, non-public
+// replacement without communicating.
+func RatchetKey(prev wgtypes.Key) wgtypes.Key {
+	input := make([]byte, 0, len(ratchetLabel)+len(prev))
+	input = append(input, ratchetLabel...)
+	input = append(input, prev[:]...)
+	return sha256.Sum256(input)
+}
+
 // DeterministicSeedKey derives a 32-byte WireGuard preshared key from a pair
 // of peer public keys. Both peers, given the same key pair, produce the same
 // output regardless of which side runs the function: the inputs are ordered