Merge pull request #4749 from matrix-org/erikj/replication_connection_backoff

Fix tightloop over connecting to replication server

1 files changed, 17 insertions, 0 deletions

diff --git a/synapse/replication/tcp/protocol.py b/synapse/replication/tcp/protocol.py
index 429471c345..49ae5b3355 100644
--- a/synapse/replication/tcp/protocol.py
+++ b/synapse/replication/tcp/protocol.py
@@ -526,6 +526,11 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
         self.server_name = server_name
         self.handler = handler
 
+        # Set of stream names that have been subscribe to, but haven't yet
+        # caught up with. This is used to track when the client has been fully
+        # connected to the remote.
+        self.streams_connecting = set()
+
         # Map of stream to batched updates. See RdataCommand for info on how
         # batching works.
         self.pending_batches = {}
@@ -548,6 +553,10 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
         # We've now finished connecting to so inform the client handler
         self.handler.update_connection(self)
 
+        # This will happen if we don't actually subscribe to any streams
+        if not self.streams_connecting:
+            self.handler.finished_connecting()
+
     def on_SERVER(self, cmd):
         if cmd.data != self.server_name:
             logger.error("[%s] Connected to wrong remote: %r", self.id(), cmd.data)
@@ -577,6 +586,12 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
             return self.handler.on_rdata(stream_name, cmd.token, rows)
 
     def on_POSITION(self, cmd):
+        # When we get a `POSITION` command it means we've finished getting
+        # missing updates for the given stream, and are now up to date.
+        self.streams_connecting.discard(cmd.stream_name)
+        if not self.streams_connecting:
+            self.handler.finished_connecting()
+
         return self.handler.on_position(cmd.stream_name, cmd.token)
 
     def on_SYNC(self, cmd):
@@ -593,6 +608,8 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
             self.id(), stream_name, token
         )
 
+        self.streams_connecting.add(stream_name)
+
         self.send_command(ReplicateCommand(stream_name, token))
 
     def on_connection_closed(self):