feat: make ServerTailnet set peers lost when it reconnects to the coordinator (#11682)

Adds support to `ServerTailnet` to set all peers lost before attempting to reconnect to the coordinator. In practice, this only really affects `wsproxy` since coderd has a local connection to the coordinator that only goes down if we're shutting down or change licenses.
This commit is contained in:
Spike Curtis
2024-01-23 13:17:56 +04:00
committed by GitHub
parent f86186eef2
commit eb12fd7d92
7 changed files with 315 additions and 7 deletions
+11 -4
View File
@@ -95,6 +95,7 @@ func NewServerTailnet(
logger: logger,
tracer: traceProvider.Tracer(tracing.TracerName),
conn: conn,
coordinatee: conn,
getMultiAgent: getMultiAgent,
cache: cache,
agentConnectionTimes: map[uuid.UUID]time.Time{},
@@ -224,13 +225,14 @@ func (s *ServerTailnet) watchAgentUpdates() {
if !ok {
if conn.IsClosed() && s.ctx.Err() == nil {
s.logger.Warn(s.ctx, "multiagent closed, reinitializing")
s.coordinatee.SetAllPeersLost()
s.reinitCoordinator()
continue
}
return
}
err := s.conn.UpdatePeers(resp.GetPeerUpdates())
err := s.coordinatee.UpdatePeers(resp.GetPeerUpdates())
if err != nil {
if xerrors.Is(err, tailnet.ErrConnClosed) {
s.logger.Warn(context.Background(), "tailnet conn closed, exiting watchAgentUpdates", slog.Error(err))
@@ -280,9 +282,14 @@ type ServerTailnet struct {
cancel func()
derpMapUpdaterClosed chan struct{}
logger slog.Logger
tracer trace.Tracer
conn *tailnet.Conn
logger slog.Logger
tracer trace.Tracer
// in prod, these are the same, but coordinatee is a subset of Conn's
// methods which makes some tests easier.
conn *tailnet.Conn
coordinatee tailnet.Coordinatee
getMultiAgent func(context.Context) (tailnet.MultiAgentConn, error)
agentConn atomic.Pointer[tailnet.MultiAgentConn]
cache *wsconncache.Cache
+75
View File
@@ -0,0 +1,75 @@
package coderd
import (
"context"
"sync/atomic"
"testing"
"time"
"github.com/google/uuid"
"go.uber.org/mock/gomock"
"cdr.dev/slog"
"cdr.dev/slog/sloggers/slogtest"
"github.com/coder/coder/v2/tailnet"
"github.com/coder/coder/v2/tailnet/tailnettest"
"github.com/coder/coder/v2/testutil"
)
// TestServerTailnet_Reconnect tests that ServerTailnet calls SetAllPeersLost on the Coordinatee
// (tailnet.Conn in production) when it disconnects from the Coordinator (via MultiAgentConn) and
// reconnects.
func TestServerTailnet_Reconnect(t *testing.T) {
t.Parallel()
logger := slogtest.Make(t, &slogtest.Options{IgnoreErrors: true}).Leveled(slog.LevelDebug)
ctrl := gomock.NewController(t)
ctx := testutil.Context(t, testutil.WaitShort)
mMultiAgent0 := tailnettest.NewMockMultiAgentConn(ctrl)
mMultiAgent1 := tailnettest.NewMockMultiAgentConn(ctrl)
mac := make(chan tailnet.MultiAgentConn, 2)
mac <- mMultiAgent0
mac <- mMultiAgent1
mCoord := tailnettest.NewMockCoordinatee(ctrl)
uut := &ServerTailnet{
ctx: ctx,
logger: logger,
coordinatee: mCoord,
getMultiAgent: func(ctx context.Context) (tailnet.MultiAgentConn, error) {
select {
case <-ctx.Done():
return nil, ctx.Err()
case m := <-mac:
return m, nil
}
},
agentConn: atomic.Pointer[tailnet.MultiAgentConn]{},
agentConnectionTimes: make(map[uuid.UUID]time.Time),
}
// reinit the Coordinator once, to load mMultiAgent0
uut.reinitCoordinator()
mMultiAgent0.EXPECT().NextUpdate(gomock.Any()).
Times(1).
Return(nil, false) // this indicates there are no more updates
closed0 := mMultiAgent0.EXPECT().IsClosed().
Times(1).
Return(true) // this triggers reconnect
setLost := mCoord.EXPECT().SetAllPeersLost().Times(1).After(closed0)
mMultiAgent1.EXPECT().NextUpdate(gomock.Any()).
Times(1).
After(setLost).
Return(nil, false)
mMultiAgent1.EXPECT().IsClosed().
Times(1).
Return(false) // this causes us to exit and not reconnect
done := make(chan struct{})
go func() {
uut.watchAgentUpdates()
close(done)
}()
testutil.RequireRecvCtx(ctx, t, done)
}