reliability fixes and disconnect-all command

This commit is contained in:
jnfrati
2026-07-03 15:26:13 +02:00
parent 520370a8b0
commit 1428970a24
22 changed files with 1394 additions and 243 deletions
+14 -11
View File
@@ -21,8 +21,11 @@ import (
)
const (
staticClientDashboard = "netbird-dashboard"
staticClientCLI = "netbird-cli"
StaticClientDashboard = "netbird-dashboard"
StaticClientCLI = "netbird-cli"
DefaultTOTPAuthenticatorID = "default-totp"
LocalConnectorID = dex.LocalConnectorID
defaultCLIRedirectURL1 = "http://localhost:53000/"
defaultCLIRedirectURL2 = "http://localhost:54000/"
defaultScopes = "openid profile email groups"
@@ -185,14 +188,14 @@ func (c *EmbeddedIdPConfig) ToYAMLConfig() (*dex.YAMLConfig, error) {
EnablePasswordDB: true,
StaticClients: []storage.Client{
{
ID: staticClientDashboard,
ID: StaticClientDashboard,
Name: "NetBird Dashboard",
Public: true,
RedirectURIs: redirectURIs,
PostLogoutRedirectURIs: sanitizePostLogoutRedirectURIs(dashboardPostLogoutRedirectURIs),
},
{
ID: staticClientCLI,
ID: StaticClientCLI,
Name: "NetBird CLI",
Public: true,
RedirectURIs: redirectURIs,
@@ -254,13 +257,13 @@ func sanitizePostLogoutRedirectURIs(uris []string) []string {
func configureMFA(cfg *dex.YAMLConfig, sessionMaxLifetime, sessionIdleTimeout string, rememberMe bool, sessionCookieEncryptionKey string) error {
cfg.MFA.Authenticators = []dex.MFAAuthenticator{{
ID: "default-totp",
ID: DefaultTOTPAuthenticatorID,
// Has to be caps otherwise it will fail
Type: "TOTP",
Config: map[string]interface{}{
"issuer": "NetBird",
},
ConnectorTypes: []string{"local"},
ConnectorTypes: []string{LocalConnectorID},
}}
if sessionMaxLifetime == "" {
@@ -736,7 +739,7 @@ func (m *EmbeddedIdPManager) GetDefaultScopes() string {
// GetCLIClientID returns the client ID for CLI authentication.
func (m *EmbeddedIdPManager) GetCLIClientID() string {
return staticClientCLI
return StaticClientCLI
}
// GetCLIRedirectURLs returns the redirect URLs configured for the CLI client.
@@ -775,7 +778,7 @@ func (m *EmbeddedIdPManager) GetLocalKeysLocation() string {
// GetClientIDs returns the OAuth2 client IDs configured for this provider.
func (m *EmbeddedIdPManager) GetClientIDs() []string {
return []string{staticClientDashboard, staticClientCLI}
return []string{StaticClientDashboard, StaticClientCLI}
}
// GetUserIDClaim returns the JWT claim name used for user identification.
@@ -792,11 +795,11 @@ func (m *EmbeddedIdPManager) IsLocalAuthDisabled() bool {
func (m *EmbeddedIdPManager) SetMFAEnabled(ctx context.Context, enabled bool) error {
var mfaChain []string
if enabled {
mfaChain = []string{"default-totp"}
mfaChain = []string{DefaultTOTPAuthenticatorID}
}
if err := m.provider.SetClientsMFAChain(ctx, []string{
staticClientCLI,
staticClientDashboard,
StaticClientCLI,
StaticClientDashboard,
}, mfaChain); err != nil {
return fmt.Errorf("failed to set MFA enabled=%v: %w", enabled, err)
}
+1 -1
View File
@@ -331,7 +331,7 @@ func TestEmbeddedIdPConfig_ToYAMLConfig_IncludesDeviceCallbackRedirectURI(t *tes
var cliRedirectURIs []string
for _, client := range yamlConfig.StaticClients {
if client.ID == staticClientCLI {
if client.ID == StaticClientCLI {
cliRedirectURIs = client.RedirectURIs
break
}
+36 -1
View File
@@ -5767,6 +5767,37 @@ func (s *SqlStore) DisconnectProxy(ctx context.Context, proxyID, sessionID strin
return nil
}
// GetAllProxies returns all reverse proxy instance rows.
func (s *SqlStore) GetAllProxies(ctx context.Context) ([]*proxy.Proxy, error) {
var proxies []*proxy.Proxy
result := s.db.Order("cluster_address, id").Find(&proxies)
if result.Error != nil {
log.WithContext(ctx).Errorf("failed to get proxies: %v", result.Error)
return nil, status.Errorf(status.Internal, "failed to get proxies")
}
return proxies, nil
}
// DisconnectAllProxies force-marks every proxy that is not already disconnected
// as disconnected, regardless of session ID. Unlike DisconnectProxy it is not
// session-guarded: it is an administrative repair helper, not part of the
// connection lifecycle. last_seen is left untouched so the stale-proxy reaper
// keeps working off the real last heartbeat. Returns the number of proxies updated.
func (s *SqlStore) DisconnectAllProxies(ctx context.Context) (int64, error) {
result := s.db.
Model(&proxy.Proxy{}).
Where("status != ?", proxy.StatusDisconnected).
Updates(map[string]any{
"status": proxy.StatusDisconnected,
"disconnected_at": time.Now(),
})
if result.Error != nil {
log.WithContext(ctx).Errorf("failed to disconnect all proxies: %v", result.Error)
return 0, status.Errorf(status.Internal, "failed to disconnect all proxies")
}
return result.RowsAffected, nil
}
// UpdateProxyHeartbeat updates the last_seen timestamp for the proxy's current session.
func (s *SqlStore) UpdateProxyHeartbeat(ctx context.Context, p *proxy.Proxy) error {
now := time.Now()
@@ -5774,7 +5805,11 @@ func (s *SqlStore) UpdateProxyHeartbeat(ctx context.Context, p *proxy.Proxy) err
result := s.db.
Model(&proxy.Proxy{}).
Where("id = ? AND session_id = ?", p.ID, p.SessionID).
Update("last_seen", now)
Updates(map[string]any{
"last_seen": now,
"status": proxy.StatusConnected,
"disconnected_at": nil,
})
if result.Error != nil {
log.WithContext(ctx).Errorf("failed to update proxy heartbeat: %v", result.Error)
@@ -0,0 +1,156 @@
package store
import (
"context"
"os"
"runtime"
"testing"
"time"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
rpproxy "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy"
)
// TestSqlStore_DisconnectAllProxies guards the administrative
// force-disconnect helper:
//
// 1. Every proxy that is not already disconnected is marked
// disconnected regardless of its session ID (unlike
// DisconnectProxy, which is session-guarded).
// 2. Rows that are already disconnected are left untouched, so their
// original disconnected_at is preserved and the returned count
// reflects only the rows that actually changed.
// 3. last_seen is not modified — the stale-proxy reaper keeps working
// off the real last heartbeat.
func TestSqlStore_DisconnectAllProxies(t *testing.T) {
if (os.Getenv("CI") == "true" && runtime.GOOS == "darwin") || runtime.GOOS == "windows" {
t.Skip("skip CI tests on darwin and windows")
}
runTestForAllEngines(t, "", func(t *testing.T, store Store) {
ctx := context.Background()
lastSeenFresh := time.Now().Add(-30 * time.Second)
lastSeenStale := time.Now().Add(-30 * time.Minute)
oldDisconnectedAt := time.Now().Add(-time.Hour)
accountID := "acct-disconnect"
proxies := []*rpproxy.Proxy{
{
ID: "p-connected-fresh",
SessionID: "sess-1",
ClusterAddress: "cluster-a.example.com",
IPAddress: "10.0.0.1",
LastSeen: lastSeenFresh,
Status: rpproxy.StatusConnected,
},
{
ID: "p-connected-stale",
SessionID: "sess-2",
ClusterAddress: "cluster-b.example.com",
IPAddress: "10.0.0.2",
AccountID: &accountID,
LastSeen: lastSeenStale,
Status: rpproxy.StatusConnected,
},
{
ID: "p-already-disconnected",
SessionID: "sess-3",
ClusterAddress: "cluster-a.example.com",
IPAddress: "10.0.0.3",
LastSeen: lastSeenStale,
Status: rpproxy.StatusDisconnected,
DisconnectedAt: &oldDisconnectedAt,
},
}
for _, p := range proxies {
require.NoError(t, store.SaveProxy(ctx, p))
}
all, err := store.GetAllProxies(ctx)
require.NoError(t, err)
require.Len(t, all, 3)
disconnected, err := store.DisconnectAllProxies(ctx)
require.NoError(t, err)
assert.Equal(t, int64(2), disconnected)
all, err = store.GetAllProxies(ctx)
require.NoError(t, err)
require.Len(t, all, 3)
byID := make(map[string]*rpproxy.Proxy, len(all))
for _, p := range all {
byID[p.ID] = p
}
for id, p := range byID {
assert.Equal(t, rpproxy.StatusDisconnected, p.Status, "proxy %s should be disconnected", id)
require.NotNil(t, p.DisconnectedAt, "proxy %s should have disconnected_at set", id)
}
// force-marked rows carry a fresh disconnected_at; the untouched row keeps its original one
assert.WithinDuration(t, time.Now(), *byID["p-connected-fresh"].DisconnectedAt, 10*time.Second)
assert.WithinDuration(t, time.Now(), *byID["p-connected-stale"].DisconnectedAt, 10*time.Second)
assert.WithinDuration(t, oldDisconnectedAt, *byID["p-already-disconnected"].DisconnectedAt, time.Second)
// last_seen is preserved so the stale reaper schedule is unaffected
assert.WithinDuration(t, lastSeenFresh, byID["p-connected-fresh"].LastSeen, time.Second)
assert.WithinDuration(t, lastSeenStale, byID["p-connected-stale"].LastSeen, time.Second)
// idempotent: a second run has nothing left to update
disconnected, err = store.DisconnectAllProxies(ctx)
require.NoError(t, err)
assert.Equal(t, int64(0), disconnected)
})
}
func TestSqlStore_UpdateProxyHeartbeatRestoresDisconnectedCurrentSession(t *testing.T) {
if (os.Getenv("CI") == "true" && runtime.GOOS == "darwin") || runtime.GOOS == "windows" {
t.Skip("skip CI tests on darwin and windows")
}
runTestForAllEngines(t, "", func(t *testing.T, store Store) {
ctx := context.Background()
proxy := &rpproxy.Proxy{
ID: "p-heartbeat",
SessionID: "sess-heartbeat",
ClusterAddress: "cluster-heartbeat.example.com",
IPAddress: "10.0.0.10",
LastSeen: time.Now().Add(-30 * time.Second),
Status: rpproxy.StatusConnected,
}
require.NoError(t, store.SaveProxy(ctx, proxy))
disconnected, err := store.DisconnectAllProxies(ctx)
require.NoError(t, err)
require.Equal(t, int64(1), disconnected)
require.NoError(t, store.UpdateProxyHeartbeat(ctx, &rpproxy.Proxy{ID: proxy.ID, SessionID: proxy.SessionID}))
all, err := store.GetAllProxies(ctx)
require.NoError(t, err)
require.Len(t, all, 1)
assert.Equal(t, rpproxy.StatusConnected, all[0].Status)
assert.Nil(t, all[0].DisconnectedAt)
assert.WithinDuration(t, time.Now(), all[0].LastSeen, 10*time.Second)
addresses, err := store.GetActiveProxyClusterAddresses(ctx)
require.NoError(t, err)
assert.Contains(t, addresses, proxy.ClusterAddress)
})
}
func TestSqlStore_GetAllProxies_Empty(t *testing.T) {
if (os.Getenv("CI") == "true" && runtime.GOOS == "darwin") || runtime.GOOS == "windows" {
t.Skip("skip CI tests on darwin and windows")
}
runTestForAllEngines(t, "", func(t *testing.T, store Store) {
all, err := store.GetAllProxies(context.Background())
require.NoError(t, err)
assert.Empty(t, all)
})
}
+2
View File
@@ -317,6 +317,8 @@ type Store interface {
GetClusterSupportsCrowdSec(ctx context.Context, clusterAddr string) *bool
GetClusterSupportsPrivate(ctx context.Context, clusterAddr string) *bool
CleanupStaleProxies(ctx context.Context, inactivityDuration time.Duration) error
GetAllProxies(ctx context.Context) ([]*proxy.Proxy, error)
DisconnectAllProxies(ctx context.Context) (int64, error)
GetProxyByAccountID(ctx context.Context, accountID string) (*proxy.Proxy, error)
CountProxiesByAccountID(ctx context.Context, accountID string) (int64, error)
IsClusterAddressConflicting(ctx context.Context, clusterAddress, accountID string) (bool, error)
+75 -45
View File
@@ -745,6 +745,21 @@ func (mr *MockStoreMockRecorder) DeleteZoneDNSRecords(ctx, accountID, zoneID int
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "DeleteZoneDNSRecords", reflect.TypeOf((*MockStore)(nil).DeleteZoneDNSRecords), ctx, accountID, zoneID)
}
// DisconnectAllProxies mocks base method.
func (m *MockStore) DisconnectAllProxies(ctx context.Context) (int64, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "DisconnectAllProxies", ctx)
ret0, _ := ret[0].(int64)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// DisconnectAllProxies indicates an expected call of DisconnectAllProxies.
func (mr *MockStoreMockRecorder) DisconnectAllProxies(ctx interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "DisconnectAllProxies", reflect.TypeOf((*MockStore)(nil).DisconnectAllProxies), ctx)
}
// DisconnectProxy mocks base method.
func (m *MockStore) DisconnectProxy(ctx context.Context, proxyID, sessionID string) error {
m.ctrl.T.Helper()
@@ -1389,6 +1404,21 @@ func (mr *MockStoreMockRecorder) GetAllEphemeralPeers(ctx, lockStrength interfac
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetAllEphemeralPeers", reflect.TypeOf((*MockStore)(nil).GetAllEphemeralPeers), ctx, lockStrength)
}
// GetAllProxies mocks base method.
func (m *MockStore) GetAllProxies(ctx context.Context) ([]*proxy.Proxy, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetAllProxies", ctx)
ret0, _ := ret[0].([]*proxy.Proxy)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetAllProxies indicates an expected call of GetAllProxies.
func (mr *MockStoreMockRecorder) GetAllProxies(ctx interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetAllProxies", reflect.TypeOf((*MockStore)(nil).GetAllProxies), ctx)
}
// GetAllProxyAccessTokens mocks base method.
func (m *MockStore) GetAllProxyAccessTokens(ctx context.Context, lockStrength LockingStrength) ([]*types2.ProxyAccessToken, error) {
m.ctrl.T.Helper()
@@ -1521,6 +1551,21 @@ func (mr *MockStoreMockRecorder) GetDNSRecordByID(ctx, lockStrength, accountID,
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetDNSRecordByID", reflect.TypeOf((*MockStore)(nil).GetDNSRecordByID), ctx, lockStrength, accountID, zoneID, recordID)
}
// GetEmbeddedProxyPeerIDsByCluster mocks base method.
func (m *MockStore) GetEmbeddedProxyPeerIDsByCluster(ctx context.Context, accountID string) (map[string][]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetEmbeddedProxyPeerIDsByCluster", ctx, accountID)
ret0, _ := ret[0].(map[string][]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetEmbeddedProxyPeerIDsByCluster indicates an expected call of GetEmbeddedProxyPeerIDsByCluster.
func (mr *MockStoreMockRecorder) GetEmbeddedProxyPeerIDsByCluster(ctx, accountID interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetEmbeddedProxyPeerIDsByCluster", reflect.TypeOf((*MockStore)(nil).GetEmbeddedProxyPeerIDsByCluster), ctx, accountID)
}
// GetExpiredEphemeralServices mocks base method.
func (m *MockStore) GetExpiredEphemeralServices(ctx context.Context, ttl time.Duration, limit int) ([]*service.Service, error) {
m.ctrl.T.Helper()
@@ -1566,6 +1611,21 @@ func (mr *MockStoreMockRecorder) GetGroupByName(ctx, lockStrength, accountID, gr
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetGroupByName", reflect.TypeOf((*MockStore)(nil).GetGroupByName), ctx, lockStrength, accountID, groupName)
}
// GetGroupIDsByPeerIDs mocks base method.
func (m *MockStore) GetGroupIDsByPeerIDs(ctx context.Context, accountID string, peerIDs []string) ([]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetGroupIDsByPeerIDs", ctx, accountID, peerIDs)
ret0, _ := ret[0].([]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetGroupIDsByPeerIDs indicates an expected call of GetGroupIDsByPeerIDs.
func (mr *MockStoreMockRecorder) GetGroupIDsByPeerIDs(ctx, accountID, peerIDs interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetGroupIDsByPeerIDs", reflect.TypeOf((*MockStore)(nil).GetGroupIDsByPeerIDs), ctx, accountID, peerIDs)
}
// GetGroupsByIDs mocks base method.
func (m *MockStore) GetGroupsByIDs(ctx context.Context, lockStrength LockingStrength, accountID string, groupIDs []string) (map[string]*types2.Group, error) {
m.ctrl.T.Helper()
@@ -1850,6 +1910,21 @@ func (mr *MockStoreMockRecorder) GetPeerIDByKey(ctx, lockStrength, key interface
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetPeerIDByKey", reflect.TypeOf((*MockStore)(nil).GetPeerIDByKey), ctx, lockStrength, key)
}
// GetPeerIDsByGroups mocks base method.
func (m *MockStore) GetPeerIDsByGroups(ctx context.Context, accountID string, groupIDs []string) ([]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetPeerIDsByGroups", ctx, accountID, groupIDs)
ret0, _ := ret[0].([]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetPeerIDsByGroups indicates an expected call of GetPeerIDsByGroups.
func (mr *MockStoreMockRecorder) GetPeerIDsByGroups(ctx, accountID, groupIDs interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetPeerIDsByGroups", reflect.TypeOf((*MockStore)(nil).GetPeerIDsByGroups), ctx, accountID, groupIDs)
}
// GetPeerIdByLabel mocks base method.
func (m *MockStore) GetPeerIdByLabel(ctx context.Context, lockStrength LockingStrength, accountID, hostname string) (string, error) {
m.ctrl.T.Helper()
@@ -1925,51 +2000,6 @@ func (mr *MockStoreMockRecorder) GetPeersByGroupIDs(ctx, accountID, groupIDs int
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetPeersByGroupIDs", reflect.TypeOf((*MockStore)(nil).GetPeersByGroupIDs), ctx, accountID, groupIDs)
}
// GetPeerIDsByGroups mocks base method.
func (m *MockStore) GetPeerIDsByGroups(ctx context.Context, accountID string, groupIDs []string) ([]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetPeerIDsByGroups", ctx, accountID, groupIDs)
ret0, _ := ret[0].([]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetPeerIDsByGroups indicates an expected call of GetPeerIDsByGroups.
func (mr *MockStoreMockRecorder) GetPeerIDsByGroups(ctx, accountID, groupIDs interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetPeerIDsByGroups", reflect.TypeOf((*MockStore)(nil).GetPeerIDsByGroups), ctx, accountID, groupIDs)
}
// GetGroupIDsByPeerIDs mocks base method.
func (m *MockStore) GetGroupIDsByPeerIDs(ctx context.Context, accountID string, peerIDs []string) ([]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetGroupIDsByPeerIDs", ctx, accountID, peerIDs)
ret0, _ := ret[0].([]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetGroupIDsByPeerIDs indicates an expected call of GetGroupIDsByPeerIDs.
func (mr *MockStoreMockRecorder) GetGroupIDsByPeerIDs(ctx, accountID, peerIDs interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetGroupIDsByPeerIDs", reflect.TypeOf((*MockStore)(nil).GetGroupIDsByPeerIDs), ctx, accountID, peerIDs)
}
// GetEmbeddedProxyPeerIDsByCluster mocks base method.
func (m *MockStore) GetEmbeddedProxyPeerIDsByCluster(ctx context.Context, accountID string) (map[string][]string, error) {
m.ctrl.T.Helper()
ret := m.ctrl.Call(m, "GetEmbeddedProxyPeerIDsByCluster", ctx, accountID)
ret0, _ := ret[0].(map[string][]string)
ret1, _ := ret[1].(error)
return ret0, ret1
}
// GetEmbeddedProxyPeerIDsByCluster indicates an expected call of GetEmbeddedProxyPeerIDsByCluster.
func (mr *MockStoreMockRecorder) GetEmbeddedProxyPeerIDsByCluster(ctx, accountID interface{}) *gomock.Call {
mr.mock.ctrl.T.Helper()
return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "GetEmbeddedProxyPeerIDsByCluster", reflect.TypeOf((*MockStore)(nil).GetEmbeddedProxyPeerIDsByCluster), ctx, accountID)
}
// GetPeersByIDs mocks base method.
func (m *MockStore) GetPeersByIDs(ctx context.Context, lockStrength LockingStrength, accountID string, peerIDs []string) (map[string]*peer.Peer, error) {
m.ctrl.T.Helper()