This PR adds supports for the WireGuard userspace implementation
using Bind interface from wireguard-go.
The newly introduced ICEBind struct implements Bind with UDPMux-based
structs from pion/ice to handle hole punching using ICE.
The core implementation was taken from StdBind of wireguard-go.
The result is a single WireGuard port that is used for host and server reflexive candidates.
Relay candidates are still handled separately and will be integrated in the following PRs.
ICEBind checks the incoming packets for being STUN or WireGuard ones
and routes them to UDPMux (to handle hole punching) or to WireGuard respectively.
When peer login expires, all remote peers are updated to exclude the peer from connecting.
Once a peer re-authenticates, the remote peers are not updated.
This peer fixes the behavior.
The peer login expiration ACL check introduced in #714
filters out peers that are expired and agents receive a network map
without that expired peers.
However, the agents should see those peers in status "Disconnected".
This PR extends the Agent <-> Management protocol
by introducing a new field OfflinePeers
that contain expired peers. Agents keep track of those and display
then just in the Status response.
The Management gRPC API has too much business logic
happening while it has to be in the Account manager.
This also needs to make more requests to the store
through the account manager.
Bug 1: When calculating the network map, peers added by a setup key
were falling under expiration logic while they shouldn't.
Bug 2: Peers HTTP API didn't return expired peers for non-admin users
because of the expired peer check in the ACL logic.
The fix applies peer expiration checks outside of the ACL logic.
When we delete a peer from an account, we save the account in the file store.
The file store maintains peerID -> accountID and peerKey -> accountID indices.
Those can't be updated when we delete a peer because the store saves the whole account
without a peer already and has no access to the removed peer.
In this PR, we dynamically check if there are stale indices when GetAccountByPeerPubKey
and GetAccountByPeerID.
Goals:
Enable peer login expiration when adding new peer
Expire peer's login when the time comes
The account manager triggers peer expiration routine in future if the
following conditions are true:
peer expiration is enabled for the account
there is at least one peer that has expiration enabled and is connected
The time of the next expiration check is based on the nearest peer expiration.
Account manager finds a peer with the oldest last login (auth) timestamp and
calculates the time when it has to run the routine as a sum of the configured
peer login expiration duration and the peer's last login time.
When triggered, the expiration routine checks whether there are expired peers.
The management server closes the update channel of these peers and updates
network map of other peers to exclude expired peers so that the expired peers
are not able to connect anywhere.
The account manager can reschedule or cancel peer expiration in the following cases:
when admin changes account setting (peer expiration enable/disable)
when admin updates the expiration duration of the account
when admin updates peer expiration (enable/disable)
when peer connects (Sync)
P.S. The network map calculation was updated to exclude peers that have login expired.
Extend HTTP API with Account endpoints to configure global peer login expiration.
GET /api/accounts
PUT /api/account/{id}/
The GET endpoint returns an array of accounts with
always one account in the list. No exceptions.
The PUT endpoint updates account settings:
PeerLoginExpiration and PeerLoginExpirationEnabled.
PeerLoginExpiration is a duration in seconds after which peers' logins will expire.
This PR adds a peer login expiration logic that requires
peers created by a user to re-authenticate (re-login) after
a certain threshold of time (24h by default).
The Account object now has a PeerLoginExpiration
property that indicates the duration after which a peer's
login will expire and a login will be required. Defaults to 24h.
There are two new properties added to the Peer object:
LastLogin that indicates the last time peer successfully used
the Login gRPC endpoint and LoginExpirationEnabled that
enables/disables peer login expiration.
The login expiration logic applies only to peers that were created
by a user and not those that were added with a setup key.
Replace Peer.Key as internal identifier with a randomly generated Peer.ID
in the Management service.
Every group now references peers by ID instead of a public key.
Every route now references peers by ID instead of a public key.
FileStore does store.json file migration on startup by generating Peer.ID and replacing
all Peer.Key identifier references .
This PR adds system activity tracking.
The management service records events like
add/remove peer, group, rule, route, etc.
The activity events are stored in the SQLite event store
and can be queried by the HTTP API.
Add a usage_limit parameter to the API.
This limits the number of times a setup key
can be used.
usage_limit == 0 indicates the the usage is inlimited.
Due to peer reconnects when restarting the Management service,
there are lots of SaveStore operations to update peer status.
Store.SavePeerStatus stores peer status separately and the
FileStore implementation stores it in memory.
This PR simplifies Store and FileStore
by keeping just the Get and Save account methods.
The AccountManager operates mostly around
a single account, so it makes sense to fetch
the whole account object from the store.
This PR brings open-telemetry metrics to the
Management service.
The Management service exposes new HTTP endpoint
/metrics on 8081 port by default.
The port can be changed by specifying
--metrics-port PORT flag when starting the service.
This PR brings user invites logic to the Management service
via HTTP API.
The POST /users/ API endpoint creates a new user in the Idp
and then in the local storage.
Once the invited user signs ups, the account invitation is redeemed.
There are a few limitations.
This works only with an enabled IdP manager.
Users that already have a registered account can't be invited.
Support Generic OAuth 2.0 Device Authorization Grant
as per RFC specification https://www.rfc-editor.org/rfc/rfc8628.
The previous version supported only Auth0 as an IDP backend.
This implementation enables the Interactive SSO Login feature
for any IDP compatible with the specification, e.g., Keycloak.
This PR fixes a race condition that happens
when agents connect to a Signal stream, multiple
times within a short amount of time. Common on
slow and unstable internet connections.
Every time an agent establishes a new connection
to Signal, Signal creates a Stream and writes an entry
to the registry of connected peers storing the stream.
Every time an agent disconnects, Signal removes the
stream from the registry.
Due to unstable connections, the agent could detect
a broken connection, and attempt to reconnect to Signal.
Signal will override the stream, but it might detect
the old broken connection later, causing peer deregistration.
It will deregister the peer leaving the client thinking
it is still connected, rejecting any messages.
When creating TLSConfig from provided certificate file, the HTTP/2 support is not enabled.
It works with Certmanager because it adds h2 support.
We enable it the same way when creating TLSConfig from files.
All the existing agents by default connect to port 33073 of the
Management service. This value is also stored in the local config.
All the agents won't switch to the new port 443
unless explicitly specified in the config.
We want the transition to be smooth for our users, therefore
this PR adds logic to check whether the old port 33073 can be
changed to 443 and updates the config automatically.
This PR is a part of an effort to use standard ports (443 or 80) that are usually allowed by default in most of the environments.
Right now Management Service runs the Let'sEncrypt manager on port 443, HTTP API server on port 33071,
and a gRPC server on port 33073. There are three separate listeners.
This PR combines these listeners into one.
With this change, the HTTP and gRPC server runs on either 443 with TLS or 80 without TLS
by default (no --port specified).
Let's Encrypt manager always runs on port 443 if enabled.
The backward compatibility server runs on port 33073 (with TLS or without).
HTTP port 33071 is obsolete and not used anymore.
Newly installed agents will connect to port 443 by default instead of port 33073 if not specified otherwise.
Right now Signal Service runs the Let'sEncrypt manager on port 80
and a gRPC server on port 10000. There are two separate listeners.
This PR combines these listeners into one with a cmux lib.
The gRPC server runs on either 443 with TLS or 80 without TLS.
Let's Encrypt manager always runs on port 80.
This PR fixes issues with the terminal when
running netbird ssh to a remote agent.
Every session looks up a user and loads its
profile. If no user is found, the connection is rejected.
The default user is root.
The Management client will try reconnecting in case.
of network issues or non-permanent errors.
If the device was off-boarded, then the client will stop retrying.
This PR adds support for SSH access through the NetBird network
without managing SSH skeys.
NetBird client app has an embedded SSH server (Linux/Mac only)
and a netbird ssh command.
There are a few places where an account is created.
When we create a new account, there should be
some defaults set. E.g. created by and group ALL.
It makes sense to add it in one place to avoid inconsistencies.
Misha Bragin