hscontrol: cancel ephemeral GC timer when a node is deleted

API and logout deletes left the disconnect timer armed; it fired later
and logged a bogus ERR for the missing node.

hscontrol/app.go

@@ -161,8 +161,8 @@ func NewHeadscale(cfg *types.Config) (*Headscale, error) {
 	ephemeralGC := db.NewEphemeralGarbageCollector(func(ni types.NodeID) {
 		node, ok := app.state.GetNodeByID(ni)
 		if !ok {
-			log.Error().Uint64("node.id", ni.Uint64()).Msg("ephemeral node deletion failed")
-			log.Debug().Caller().Uint64("node.id", ni.Uint64()).Msg("ephemeral node deletion failed because node not found in NodeStore")
+			// A delete that raced this timer already did its job.
+			log.Debug().Caller().Uint64("node.id", ni.Uint64()).Msg("ephemeral node already deleted, skipping garbage collection")
 
 			return
 		}
@@ -1274,6 +1274,14 @@ func readOrCreatePrivateKey(path string) (*key.MachinePrivate, error) {
 // All change should be enqueued here and empty will be automatically
 // ignored.
 func (h *Headscale) Change(cs ...change.Change) {
+	// Every delete path publishes its removal here; a deleted node's
+	// inactivity timer has nothing left to collect.
+	for _, c := range cs {
+		for _, id := range c.DeletedNodes {
+			h.ephemeralGC.Cancel(id)
+		}
+	}
+
 	h.mapBatcher.AddWork(slices.Concat(cs, h.state.DrainSelfRefreshes())...)
 }
 

hscontrol/poll_test.go

@@ -424,6 +424,45 @@ func TestEphemeralGCDoesNotDeleteReconnectedNode(t *testing.T) {
 		"an idle ephemeral node must still be garbage collected")
 }
 
+// TestDeleteNodeCancelsEphemeralGC deletes a disconnected ephemeral node
+// through the API while its inactivity timer is armed. The timer belongs to
+// the deleted node, so the delete must disarm it rather than leave it to fire
+// at a node that is no longer there.
+func TestDeleteNodeCancelsEphemeralGC(t *testing.T) {
+	t.Parallel()
+
+	app := createTestApp(t)
+	app.StartEphemeralGCForTest(t)
+
+	user := app.state.CreateUserForTest("eph-gc-delete-user")
+	pak, err := app.state.CreatePreAuthKey(user.TypedID(), false, true, nil, nil)
+	require.NoError(t, err)
+
+	nodeKey := key.NewNode()
+
+	_, err = app.handleRegister(context.Background(), tailcfg.RegisterRequest{
+		Auth:     &tailcfg.RegisterResponseAuth{AuthKey: pak.Key},
+		NodeKey:  nodeKey.Public(),
+		Hostinfo: &tailcfg.Hostinfo{Hostname: "eph-gc-delete-node"},
+		Expiry:   time.Now().Add(24 * time.Hour),
+	}, key.NewMachine().Public())
+	require.NoError(t, err)
+
+	node, ok := app.state.GetNodeByNodeKey(nodeKey.Public())
+	require.True(t, ok)
+	require.True(t, node.IsEphemeral(), "test sanity: node must be ephemeral")
+
+	// The state a disconnect leaves behind.
+	app.ephemeralGC.Schedule(node.ID(), time.Hour)
+	require.True(t, app.ephemeralGC.IsScheduled(node.ID()), "test sanity: GC timer must be armed")
+
+	api := registerAPIV2(t, app)
+	require.Equal(t, http.StatusOK, api.Delete("/api/v2/device/"+node.ID().String()).Code)
+
+	assert.False(t, app.ephemeralGC.IsScheduled(node.ID()),
+		"deleting a node must cancel its ephemeral GC timer")
+}
+
 // TestGitHubIssue3129_TransientlyBlockedWriteDoesNotLeaveLiveStaleSession
 // tests the scenario reported in
 // https://github.com/juanfont/headscale/issues/3129.