Merge pull request #156 from matrix-org/erikj/join_perf

Make joining #matrix:matrix.org over federation quicker

3 files changed, 22 insertions, 12 deletions

diff --git a/synapse/federation/federation_base.py b/synapse/federation/federation_base.py
index 5217d91aab..f0430b2cb1 100644
--- a/synapse/federation/federation_base.py
+++ b/synapse/federation/federation_base.py
@@ -80,6 +80,7 @@ class FederationBase(object):
                             destinations=[pdu.origin],
                             event_id=pdu.event_id,
                             outlier=outlier,
+                            timeout=10000,
                         )
 
                         if new_pdu:
diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index 3a7bc0c9a7..3249060bcf 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -22,6 +22,7 @@ from .units import Edu
 from synapse.api.errors import (
     CodeMessageException, HttpResponseException, SynapseError,
 )
+from synapse.util import unwrapFirstError
 from synapse.util.expiringcache import ExpiringCache
 from synapse.util.logutils import log_function
 from synapse.events import FrozenEvent
@@ -173,7 +174,7 @@ class FederationClient(FederationBase):
 
     @defer.inlineCallbacks
     @log_function
-    def get_pdu(self, destinations, event_id, outlier=False):
+    def get_pdu(self, destinations, event_id, outlier=False, timeout=None):
         """Requests the PDU with given origin and ID from the remote home
         servers.
 
@@ -189,6 +190,8 @@ class FederationClient(FederationBase):
             outlier (bool): Indicates whether the PDU is an `outlier`, i.e. if
                 it's from an arbitary point in the context as opposed to part
                 of the current block of PDUs. Defaults to `False`
+            timeout (int): How long to try (in ms) each destination for before
+                moving to the next destination. None indicates no timeout.
 
         Returns:
             Deferred: Results in the requested PDU.
@@ -212,7 +215,7 @@ class FederationClient(FederationBase):
 
                 with limiter:
                     transaction_data = yield self.transport_layer.get_event(
-                        destination, event_id
+                        destination, event_id, timeout=timeout,
                     )
 
                     logger.debug("transaction_data %r", transaction_data)
@@ -370,13 +373,17 @@ class FederationClient(FederationBase):
                     for p in content.get("auth_chain", [])
                 ]
 
-                signed_state = yield self._check_sigs_and_hash_and_fetch(
-                    destination, state, outlier=True
-                )
-
-                signed_auth = yield self._check_sigs_and_hash_and_fetch(
-                    destination, auth_chain, outlier=True
-                )
+                signed_state, signed_auth = yield defer.gatherResults(
+                    [
+                        self._check_sigs_and_hash_and_fetch(
+                            destination, state, outlier=True
+                        ),
+                        self._check_sigs_and_hash_and_fetch(
+                            destination, auth_chain, outlier=True
+                        )
+                    ],
+                    consumeErrors=True
+                ).addErrback(unwrapFirstError)
 
                 auth_chain.sort(key=lambda e: e.depth)
 
@@ -518,7 +525,7 @@ class FederationClient(FederationBase):
             # Are we missing any?
 
             seen_events = set(earliest_events_ids)
-            seen_events.update(e.event_id for e in signed_events)
+            seen_events.update(e.event_id for e in signed_events if e)
 
             missing_events = {}
             for e in itertools.chain(latest_events, signed_events):
diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py
index 80d03012b7..610a4c3163 100644
--- a/synapse/federation/transport/client.py
+++ b/synapse/federation/transport/client.py
@@ -50,13 +50,15 @@ class TransportLayerClient(object):
         )
 
     @log_function
-    def get_event(self, destination, event_id):
+    def get_event(self, destination, event_id, timeout=None):
         """ Requests the pdu with give id and origin from the given server.
 
         Args:
             destination (str): The host name of the remote home server we want
                 to get the state from.
             event_id (str): The id of the event being requested.
+            timeout (int): How long to try (in ms) the destination for before
+                giving up. None indicates no timeout.
 
         Returns:
             Deferred: Results in a dict received from the remote homeserver.
@@ -65,7 +67,7 @@ class TransportLayerClient(object):
                      destination, event_id)
 
         path = PREFIX + "/event/%s/" % (event_id, )
-        return self.client.get_json(destination, path=path)
+        return self.client.get_json(destination, path=path, timeout=timeout)
 
     @log_function
     def backfill(self, destination, room_id, event_tuples, limit):