bridgev2/config: add limit for unknown error auto-reconnects

bridgev2/bridgeconfig/config.go

@@ -62,29 +62,30 @@ type CleanupOnLogouts struct {
 }
 
 type BridgeConfig struct {
-	CommandPrefix             string           `yaml:"command_prefix"`
-	PersonalFilteringSpaces   bool             `yaml:"personal_filtering_spaces"`
-	PrivateChatPortalMeta     bool             `yaml:"private_chat_portal_meta"`
-	AsyncEvents               bool             `yaml:"async_events"`
-	SplitPortals              bool             `yaml:"split_portals"`
-	ResendBridgeInfo          bool             `yaml:"resend_bridge_info"`
-	NoBridgeInfoStateKey      bool             `yaml:"no_bridge_info_state_key"`
-	BridgeStatusNotices       string           `yaml:"bridge_status_notices"`
-	UnknownErrorAutoReconnect time.Duration    `yaml:"unknown_error_auto_reconnect"`
-	BridgeMatrixLeave         bool             `yaml:"bridge_matrix_leave"`
-	BridgeNotices             bool             `yaml:"bridge_notices"`
-	TagOnlyOnCreate           bool             `yaml:"tag_only_on_create"`
-	OnlyBridgeTags            []event.RoomTag  `yaml:"only_bridge_tags"`
-	MuteOnlyOnCreate          bool             `yaml:"mute_only_on_create"`
-	DeduplicateMatrixMessages bool             `yaml:"deduplicate_matrix_messages"`
-	CrossRoomReplies          bool             `yaml:"cross_room_replies"`
-	OutgoingMessageReID       bool             `yaml:"outgoing_message_re_id"`
-	RevertFailedStateChanges  bool             `yaml:"revert_failed_state_changes"`
-	KickMatrixUsers           bool             `yaml:"kick_matrix_users"`
-	CleanupOnLogout           CleanupOnLogouts `yaml:"cleanup_on_logout"`
-	Relay                     RelayConfig      `yaml:"relay"`
-	Permissions               PermissionConfig `yaml:"permissions"`
-	Backfill                  BackfillConfig   `yaml:"backfill"`
+	CommandPrefix                 string           `yaml:"command_prefix"`
+	PersonalFilteringSpaces       bool             `yaml:"personal_filtering_spaces"`
+	PrivateChatPortalMeta         bool             `yaml:"private_chat_portal_meta"`
+	AsyncEvents                   bool             `yaml:"async_events"`
+	SplitPortals                  bool             `yaml:"split_portals"`
+	ResendBridgeInfo              bool             `yaml:"resend_bridge_info"`
+	NoBridgeInfoStateKey          bool             `yaml:"no_bridge_info_state_key"`
+	BridgeStatusNotices           string           `yaml:"bridge_status_notices"`
+	UnknownErrorAutoReconnect     time.Duration    `yaml:"unknown_error_auto_reconnect"`
+	UnknownErrorMaxAutoReconnects int              `yaml:"unknown_error_max_auto_reconnects"`
+	BridgeMatrixLeave             bool             `yaml:"bridge_matrix_leave"`
+	BridgeNotices                 bool             `yaml:"bridge_notices"`
+	TagOnlyOnCreate               bool             `yaml:"tag_only_on_create"`
+	OnlyBridgeTags                []event.RoomTag  `yaml:"only_bridge_tags"`
+	MuteOnlyOnCreate              bool             `yaml:"mute_only_on_create"`
+	DeduplicateMatrixMessages     bool             `yaml:"deduplicate_matrix_messages"`
+	CrossRoomReplies              bool             `yaml:"cross_room_replies"`
+	OutgoingMessageReID           bool             `yaml:"outgoing_message_re_id"`
+	RevertFailedStateChanges      bool             `yaml:"revert_failed_state_changes"`
+	KickMatrixUsers               bool             `yaml:"kick_matrix_users"`
+	CleanupOnLogout               CleanupOnLogouts `yaml:"cleanup_on_logout"`
+	Relay                         RelayConfig      `yaml:"relay"`
+	Permissions                   PermissionConfig `yaml:"permissions"`
+	Backfill                      BackfillConfig   `yaml:"backfill"`
 }
 
 type MatrixConfig struct {

bridgev2/bridgeconfig/upgrade.go

@@ -33,6 +33,7 @@ func doUpgrade(helper up.Helper) {
 	helper.Copy(up.Bool, "bridge", "no_bridge_info_state_key")
 	helper.Copy(up.Str|up.Null, "bridge", "bridge_status_notices")
 	helper.Copy(up.Str|up.Int|up.Null, "bridge", "unknown_error_auto_reconnect")
+	helper.Copy(up.Int, "bridge", "unknown_error_max_auto_reconnects")
 	helper.Copy(up.Bool, "bridge", "bridge_matrix_leave")
 	helper.Copy(up.Bool, "bridge", "bridge_notices")
 	helper.Copy(up.Bool, "bridge", "tag_only_on_create")

bridgev2/bridgestate.go

@@ -37,6 +37,8 @@ type BridgeStateQueue struct {
 
 	stopChan      chan struct{}
 	stopReconnect atomic.Pointer[context.CancelFunc]
+
+	unknownErrorReconnects int
 }
 
 func (br *Bridge) SendGlobalBridgeState(state status.BridgeState) {
@@ -192,8 +194,14 @@ func (bsq *BridgeStateQueue) unknownErrorReconnect(triggeredBy status.BridgeStat
 	} else if prevUnsent.StateEvent != status.StateUnknownError || prev.StateEvent != status.StateUnknownError {
 		log.Debug().Msg("Not reconnecting as the previous state was not an unknown error")
 		return
+	} else if bsq.unknownErrorReconnects > bsq.bridge.Config.UnknownErrorMaxAutoReconnects {
+		log.Warn().Msg("Not reconnecting as the maximum number of unknown error reconnects has been reached")
+		return
 	}
-	log.Info().Msg("Disconnecting and reconnecting login due to unknown error")
+	bsq.unknownErrorReconnects++
+	log.Info().
+		Int("reconnect_num", bsq.unknownErrorReconnects).
+		Msg("Disconnecting and reconnecting login due to unknown error")
 	bsq.login.Disconnect()
 	log.Debug().Msg("Disconnection finished, recreating client and reconnecting")
 	err := bsq.login.recreateClient(ctx)

bridgev2/matrix/mxmain/example-config.yaml

@@ -29,6 +29,9 @@ bridge:
     # How long after an unknown error should the bridge attempt a full reconnect?
     # Must be at least 1 minute. The bridge will add an extra ±20% jitter to this value.
     unknown_error_auto_reconnect: null
+    # Maximum number of times to do the auto-reconnect above.
+    # The counter is per login, but is never reset except on logout and restart.
+    unknown_error_max_auto_reconnects: 10
 
     # Should leaving Matrix rooms be bridged as leaving groups on the remote network?
     bridge_matrix_leave: false