Don't do regular federation if should be using mailserver

federationapi/internal/perform.go

@@ -24,6 +24,11 @@ func (r *FederationInternalAPI) PerformDirectoryLookup(
 	request *api.PerformDirectoryLookupRequest,
 	response *api.PerformDirectoryLookupResponse,
 ) (err error) {
+	stats := r.statistics.ForServer(request.ServerName)
+	if stats.AssumedOffline() && len(stats.KnownMailservers()) > 0 {
+		return fmt.Errorf("not performing federation since server is assumed offline with known mailboxes")
+	}
+
 	dir, err := r.federation.LookupRoomAlias(
 		ctx,
 		request.ServerName,
@@ -143,6 +148,11 @@ func (r *FederationInternalAPI) performJoinUsingServer(
 	supportedVersions []gomatrixserverlib.RoomVersion,
 	unsigned map[string]interface{},
 ) error {
+	stats := r.statistics.ForServer(serverName)
+	if stats.AssumedOffline() && len(stats.KnownMailservers()) > 0 {
+		return fmt.Errorf("not performing federation since server is assumed offline with known mailboxes")
+	}
+
 	// Try to perform a make_join using the information supplied in the
 	// request.
 	respMakeJoin, err := r.federation.MakeJoin(
@@ -398,6 +408,11 @@ func (r *FederationInternalAPI) performOutboundPeekUsingServer(
 	serverName gomatrixserverlib.ServerName,
 	supportedVersions []gomatrixserverlib.RoomVersion,
 ) error {
+	stats := r.statistics.ForServer(serverName)
+	if stats.AssumedOffline() && len(stats.KnownMailservers()) > 0 {
+		return fmt.Errorf("not performing federation since server is assumed offline with known mailboxes")
+	}
+
 	// create a unique ID for this peek.
 	// for now we just use the room ID again. In future, if we ever
 	// support concurrent peeks to the same room with different filters
@@ -501,6 +516,11 @@ func (r *FederationInternalAPI) PerformLeave(
 	// Try each server that we were provided until we land on one that
 	// successfully completes the make-leave send-leave dance.
 	for _, serverName := range request.ServerNames {
+		stats := r.statistics.ForServer(serverName)
+		if stats.AssumedOffline() && len(stats.KnownMailservers()) > 0 {
+			continue
+		}
+
 		// Try to perform a make_leave using the information supplied in the
 		// request.
 		respMakeLeave, err := r.federation.MakeLeave(
@@ -594,6 +614,11 @@ func (r *FederationInternalAPI) PerformInvite(
 		return fmt.Errorf("gomatrixserverlib.SplitID: %w", err)
 	}
 
+	stats := r.statistics.ForServer(destination)
+	if stats.AssumedOffline() && len(stats.KnownMailservers()) > 0 {
+		return fmt.Errorf("not performing federation since server is assumed offline with known mailboxes")
+	}
+
 	logrus.WithFields(logrus.Fields{
 		"event_id":     request.Event.EventID(),
 		"user_id":      *request.Event.StateKey(),

federationapi/statistics/statistics.go

@@ -65,8 +65,9 @@ func (s *Statistics) ForServer(serverName gomatrixserverlib.ServerName) *ServerS
 	if !found {
 		s.mutex.Lock()
 		server = &ServerStatistics{
-			statistics: s,
+			statistics:       s,
-			serverName: serverName,
+			serverName:       serverName,
+			knownMailservers: []gomatrixserverlib.ServerName{},
 		}
 		s.servers[serverName] = server
 		s.mutex.Unlock()
@@ -85,16 +86,17 @@ func (s *Statistics) ForServer(serverName gomatrixserverlib.ServerName) *ServerS
 // many times we failed etc. It also manages the backoff time and black-
 // listing a remote host if it remains uncooperative.
 type ServerStatistics struct {
-	statistics      *Statistics                  //
+	statistics       *Statistics                  //
-	serverName      gomatrixserverlib.ServerName //
+	serverName       gomatrixserverlib.ServerName //
-	blacklisted     atomic.Bool                  // is the node blacklisted
+	blacklisted      atomic.Bool                  // is the node blacklisted
-	assumedOffline  atomic.Bool                  // is the node assumed to be offline
+	assumedOffline   atomic.Bool                  // is the node assumed to be offline
-	backoffStarted  atomic.Bool                  // is the backoff started
+	backoffStarted   atomic.Bool                  // is the backoff started
-	backoffUntil    atomic.Value                 // time.Time until this backoff interval ends
+	backoffUntil     atomic.Value                 // time.Time until this backoff interval ends
-	backoffCount    atomic.Uint32                // number of times BackoffDuration has been called
+	backoffCount     atomic.Uint32                // number of times BackoffDuration has been called
-	successCounter  atomic.Uint32                // how many times have we succeeded?
+	successCounter   atomic.Uint32                // how many times have we succeeded?
-	backoffNotifier func()                       // notifies destination queue when backoff completes
+	backoffNotifier  func()                       // notifies destination queue when backoff completes
-	notifierMutex   sync.Mutex
+	notifierMutex    sync.Mutex
+	knownMailservers []gomatrixserverlib.ServerName
 }
 
 const maxJitterMultiplier = 1.4
@@ -245,3 +247,7 @@ func (s *ServerStatistics) RemoveBlacklist() {
 func (s *ServerStatistics) SuccessCount() uint32 {
 	return s.successCounter.Load()
 }
+
+func (s *ServerStatistics) KnownMailservers() []gomatrixserverlib.ServerName {
+	return s.knownMailservers
+}