1 files changed, 24 insertions, 7 deletions
diff --git a/synapse/app/generic_worker.py b/synapse/app/generic_worker.py
index 5afe52f8d4..f3ec2a34ec 100644
--- a/synapse/app/generic_worker.py
+++ b/synapse/app/generic_worker.py
@@ -863,9 +863,24 @@ class FederationSenderHandler(object):
         a FEDERATION_ACK back to the master, and stores the token that we have processed
          in `federation_stream_position` so that we can restart where we left off.
         """
-        try:
-            self.federation_position = token
+        self.federation_position = token
+
+        # We save and send the ACK to master asynchronously, so we don't block
+        # processing on persistence. We don't need to do this operation for
+        # every single RDATA we receive, we just need to do it periodically.
+
+        if self._fed_position_linearizer.is_queued(None):
+            # There is already a task queued up to save and send the token, so
+            # no need to queue up another task.
+            return
+
+        run_as_background_process("_save_and_send_ack", self._save_and_send_ack)
 
+    async def _save_and_send_ack(self):
+        """Save the current federation position in the database and send an ACK
+        to master with where we're up to.
+        """
+        try:
             # We linearize here to ensure we don't have races updating the token
             #
             # XXX this appears to be redundant, since the ReplicationCommandHandler
@@ -875,16 +890,18 @@ class FederationSenderHandler(object):
             # we're not being re-entered?
 
             with (await self._fed_position_linearizer.queue(None)):
+                # We persist and ack the same position, so we take a copy of it
+                # here as otherwise it can get modified from underneath us.
+                current_position = self.federation_position
+
                 await self.store.update_federation_out_pos(
-                    "federation", self.federation_position
+                    "federation", current_position
                 )
 
                 # We ACK this token over replication so that the master can drop
                 # its in memory queues
-                self._hs.get_tcp_replication().send_federation_ack(
-                    self.federation_position
-                )
-                self._last_ack = self.federation_position
+                self._hs.get_tcp_replication().send_federation_ack(current_position)
+                self._last_ack = current_position
         except Exception:
             logger.exception("Error updating federation stream position")