Merge pull request #998 from matrix-org/erikj/pdu_fail_cache

Various federation /event/ improvements

2 files changed, 53 insertions, 21 deletions

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index da95c2ad6d..9ba3151713 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -51,10 +51,34 @@ sent_edus_counter = metrics.register_counter("sent_edus")
 sent_queries_counter = metrics.register_counter("sent_queries", labels=["type"])
 
 
+PDU_RETRY_TIME_MS = 1 * 60 * 1000
+
+
 class FederationClient(FederationBase):
     def __init__(self, hs):
         super(FederationClient, self).__init__(hs)
 
+        self.pdu_destination_tried = {}
+        self._clock.looping_call(
+            self._clear_tried_cache, 60 * 1000,
+        )
+
+    def _clear_tried_cache(self):
+        """Clear pdu_destination_tried cache"""
+        now = self._clock.time_msec()
+
+        old_dict = self.pdu_destination_tried
+        self.pdu_destination_tried = {}
+
+        for event_id, destination_dict in old_dict.items():
+            destination_dict = {
+                dest: time
+                for dest, time in destination_dict.items()
+                if time + PDU_RETRY_TIME_MS > now
+            }
+            if destination_dict:
+                self.pdu_destination_tried[event_id] = destination_dict
+
     def start_get_pdu_cache(self):
         self._get_pdu_cache = ExpiringCache(
             cache_name="get_pdu_cache",
@@ -240,8 +264,15 @@ class FederationClient(FederationBase):
             if ev:
                 defer.returnValue(ev)
 
+        pdu_attempts = self.pdu_destination_tried.setdefault(event_id, {})
+
         pdu = None
         for destination in destinations:
+            now = self._clock.time_msec()
+            last_attempt = pdu_attempts.get(destination, 0)
+            if last_attempt + PDU_RETRY_TIME_MS > now:
+                continue
+
             try:
                 limiter = yield get_retry_limiter(
                     destination,
@@ -269,25 +300,19 @@ class FederationClient(FederationBase):
 
                         break
 
-            except SynapseError as e:
-                logger.info(
-                    "Failed to get PDU %s from %s because %s",
-                    event_id, destination, e,
-                )
-                continue
-            except CodeMessageException as e:
-                if 400 <= e.code < 500:
-                    raise
+                pdu_attempts[destination] = now
 
+            except SynapseError as e:
                 logger.info(
                     "Failed to get PDU %s from %s because %s",
                     event_id, destination, e,
                 )
-                continue
             except NotRetryingDestination as e:
                 logger.info(e.message)
                 continue
             except Exception as e:
+                pdu_attempts[destination] = now
+
                 logger.info(
                     "Failed to get PDU %s from %s because %s",
                     event_id, destination, e,
@@ -406,7 +431,7 @@ class FederationClient(FederationBase):
             events and the second is a list of event ids that we failed to fetch.
         """
         if return_local:
-            seen_events = yield self.store.get_events(event_ids)
+            seen_events = yield self.store.get_events(event_ids, allow_rejected=True)
             signed_events = seen_events.values()
         else:
             seen_events = yield self.store.have_events(event_ids)
diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py
index 618cb53629..ff6bb475b5 100644
--- a/synapse/handlers/federation.py
+++ b/synapse/handlers/federation.py
@@ -249,7 +249,7 @@ class FederationHandler(BaseHandler):
                         if ev.type != EventTypes.Member:
                             continue
                         try:
-                            domain = UserID.from_string(ev.state_key).domain
+                            domain = get_domain_from_id(ev.state_key)
                         except:
                             continue
 
@@ -1093,16 +1093,17 @@ class FederationHandler(BaseHandler):
         )
 
         if event:
-            # FIXME: This is a temporary work around where we occasionally
-            # return events slightly differently than when they were
-            # originally signed
-            event.signatures.update(
-                compute_event_signature(
-                    event,
-                    self.hs.hostname,
-                    self.hs.config.signing_key[0]
+            if self.hs.is_mine_id(event.event_id):
+                # FIXME: This is a temporary work around where we occasionally
+                # return events slightly differently than when they were
+                # originally signed
+                event.signatures.update(
+                    compute_event_signature(
+                        event,
+                        self.hs.hostname,
+                        self.hs.config.signing_key[0]
+                    )
                 )
-            )
 
             if do_auth:
                 in_room = yield self.auth.check_host_in_room(
@@ -1112,6 +1113,12 @@ class FederationHandler(BaseHandler):
                 if not in_room:
                     raise AuthError(403, "Host not in room.")
 
+                events = yield self._filter_events_for_server(
+                    origin, event.room_id, [event]
+                )
+
+                event = events[0]
+
             defer.returnValue(event)
         else:
             defer.returnValue(None)