Don't go into federation catch up mode so easily (#9561)

Federation catch up mode is very inefficient if the number of events
that the remote server has missed is small, since handling gaps can be
very expensive, c.f. #9492.

Instead of going into catch up mode whenever we see an error, we instead
do so only if we've backed off from trying the remote for more than an
hour (the assumption being that in such a case it is more than a
transient failure).

1 files changed, 14 insertions, 34 deletions

diff --git a/synapse/federation/sender/transaction_manager.py b/synapse/federation/sender/transaction_manager.py
index 2a9cd063c4..07b740c2f2 100644
--- a/synapse/federation/sender/transaction_manager.py
+++ b/synapse/federation/sender/transaction_manager.py
@@ -69,15 +69,12 @@ class TransactionManager:
         destination: str,
         pdus: List[EventBase],
         edus: List[Edu],
-    ) -> bool:
+    ) -> None:
         """
         Args:
             destination: The destination to send to (e.g. 'example.org')
             pdus: In-order list of PDUs to send
             edus: List of EDUs to send
-
-        Returns:
-            True iff the transaction was successful
         """
 
         # Make a transaction-sending opentracing span. This span follows on from
@@ -96,8 +93,6 @@ class TransactionManager:
                 edu.strip_context()
 
         with start_active_span_follows_from("send_transaction", span_contexts):
-            success = True
-
             logger.debug("TX [%s] _attempt_new_transaction", destination)
 
             txn_id = str(self._next_txn_id)
@@ -152,44 +147,29 @@ class TransactionManager:
                 response = await self._transport_layer.send_transaction(
                     transaction, json_data_cb
                 )
-                code = 200
             except HttpResponseException as e:
                 code = e.code
                 response = e.response
 
-                if e.code in (401, 404, 429) or 500 <= e.code:
-                    logger.info(
-                        "TX [%s] {%s} got %d response", destination, txn_id, code
-                    )
-                    raise e
-
-            logger.info("TX [%s] {%s} got %d response", destination, txn_id, code)
-
-            if code == 200:
-                for e_id, r in response.get("pdus", {}).items():
-                    if "error" in r:
-                        logger.warning(
-                            "TX [%s] {%s} Remote returned error for %s: %s",
-                            destination,
-                            txn_id,
-                            e_id,
-                            r,
-                        )
-            else:
-                for p in pdus:
+                set_tag(tags.ERROR, True)
+
+                logger.info("TX [%s] {%s} got %d response", destination, txn_id, code)
+                raise
+
+            logger.info("TX [%s] {%s} got 200 response", destination, txn_id)
+
+            for e_id, r in response.get("pdus", {}).items():
+                if "error" in r:
                     logger.warning(
-                        "TX [%s] {%s} Failed to send event %s",
+                        "TX [%s] {%s} Remote returned error for %s: %s",
                         destination,
                         txn_id,
-                        p.event_id,
+                        e_id,
+                        r,
                     )
-                success = False
 
-            if success and pdus and destination in self._federation_metrics_domains:
+            if pdus and destination in self._federation_metrics_domains:
                 last_pdu = pdus[-1]
                 last_pdu_ts_metric.labels(server_name=destination).set(
                     last_pdu.origin_server_ts / 1000
                 )
-
-            set_tag(tags.ERROR, not success)
-            return success