From f91df1f761b1e9e4da184560b0e7d9557129d064 Mon Sep 17 00:00:00 2001
From: Erik Johnston <erik@matrix.org>
Date: Wed, 10 Aug 2016 11:31:46 +0100
Subject: Store if we fail to fetch an event from a destination

---
 synapse/federation/federation_client.py | 37 ++++++++++++++++++++++++++++++++-
 1 file changed, 36 insertions(+), 1 deletion(-)

(limited to 'synapse/federation/federation_client.py')

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index da95c2ad6d..baa672c4ac 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,
@@ -276,9 +307,11 @@ class FederationClient(FederationBase):
                 )
                 continue
             except CodeMessageException as e:
-                if 400 <= e.code < 500:
+                if 400 <= e.code < 500 and e.code != 404:
                     raise
 
+                pdu_attempts[destination] = now
+
                 logger.info(
                     "Failed to get PDU %s from %s because %s",
                     event_id, destination, e,
@@ -288,6 +321,8 @@ class FederationClient(FederationBase):
                 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,
-- 
cgit 1.5.1


From ea8c4094dbaa9cec30c543a03f451d2555d1d23d Mon Sep 17 00:00:00 2001
From: Erik Johnston <erik@matrix.org>
Date: Wed, 10 Aug 2016 13:26:13 +0100
Subject: Also pull out rejected events

---
 synapse/federation/federation_client.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'synapse/federation/federation_client.py')

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index baa672c4ac..56115a87d7 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -441,7 +441,7 @@ 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)
-- 
cgit 1.5.1


From 487bc49bf8dcaadd6abd9cee1ef762f1bf0d35a7 Mon Sep 17 00:00:00 2001
From: Erik Johnston <erik@matrix.org>
Date: Wed, 10 Aug 2016 13:39:12 +0100
Subject: Don't stop on 4xx series errors

---
 synapse/federation/federation_client.py | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)

(limited to 'synapse/federation/federation_client.py')

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index 56115a87d7..9ba3151713 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -300,23 +300,13 @@ 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 and e.code != 404:
-                    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
-- 
cgit 1.5.1