Merge branch 'notifier_unify' into notifier_performance

Conflicts:
	synapse/notifier.py

1 files changed, 157 insertions, 16 deletions

diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py
index 85e2757227..7d9906039e 100644
--- a/synapse/handlers/federation.py
+++ b/synapse/handlers/federation.py
@@ -18,9 +18,11 @@
 from ._base import BaseHandler
 
 from synapse.api.errors import (
-    AuthError, FederationError, StoreError,
+    AuthError, FederationError, StoreError, CodeMessageException, SynapseError,
 )
 from synapse.api.constants import EventTypes, Membership, RejectedReason
+from synapse.util import unwrapFirstError
+from synapse.util.logcontext import PreserveLoggingContext
 from synapse.util.logutils import log_function
 from synapse.util.async import run_on_reactor
 from synapse.util.frozenutils import unfreeze
@@ -29,6 +31,8 @@ from synapse.crypto.event_signing import (
 )
 from synapse.types import UserID
 
+from synapse.util.retryutils import NotRetryingDestination
+
 from twisted.internet import defer
 
 import itertools
@@ -197,9 +201,10 @@ class FederationHandler(BaseHandler):
                 target_user = UserID.from_string(target_user_id)
                 extra_users.append(target_user)
 
-            d = self.notifier.on_new_room_event(
-                event, extra_users=extra_users
-            )
+            with PreserveLoggingContext():
+                d = self.notifier.on_new_room_event(
+                    event, extra_users=extra_users
+                )
 
             def log_failure(f):
                 logger.warn(
@@ -218,10 +223,11 @@ class FederationHandler(BaseHandler):
 
     @log_function
     @defer.inlineCallbacks
-    def backfill(self, dest, room_id, limit):
+    def backfill(self, dest, room_id, limit, extremities=[]):
         """ Trigger a backfill request to `dest` for the given `room_id`
         """
-        extremities = yield self.store.get_oldest_events_in_room(room_id)
+        if not extremities:
+            extremities = yield self.store.get_oldest_events_in_room(room_id)
 
         pdus = yield self.replication_layer.backfill(
             dest,
@@ -249,6 +255,138 @@ class FederationHandler(BaseHandler):
         defer.returnValue(events)
 
     @defer.inlineCallbacks
+    def maybe_backfill(self, room_id, current_depth):
+        """Checks the database to see if we should backfill before paginating,
+        and if so do.
+        """
+        extremities = yield self.store.get_oldest_events_with_depth_in_room(
+            room_id
+        )
+
+        if not extremities:
+            logger.debug("Not backfilling as no extremeties found.")
+            return
+
+        # Check if we reached a point where we should start backfilling.
+        sorted_extremeties_tuple = sorted(
+            extremities.items(),
+            key=lambda e: -int(e[1])
+        )
+        max_depth = sorted_extremeties_tuple[0][1]
+
+        if current_depth > max_depth:
+            logger.debug(
+                "Not backfilling as we don't need to. %d < %d",
+                max_depth, current_depth,
+            )
+            return
+
+        # Now we need to decide which hosts to hit first.
+
+        # First we try hosts that are already in the room
+        # TODO: HEURISTIC ALERT.
+
+        curr_state = yield self.state_handler.get_current_state(room_id)
+
+        def get_domains_from_state(state):
+            joined_users = [
+                (state_key, int(event.depth))
+                for (e_type, state_key), event in state.items()
+                if e_type == EventTypes.Member
+                and event.membership == Membership.JOIN
+            ]
+
+            joined_domains = {}
+            for u, d in joined_users:
+                try:
+                    dom = UserID.from_string(u).domain
+                    old_d = joined_domains.get(dom)
+                    if old_d:
+                        joined_domains[dom] = min(d, old_d)
+                    else:
+                        joined_domains[dom] = d
+                except:
+                    pass
+
+            return sorted(joined_domains.items(), key=lambda d: d[1])
+
+        curr_domains = get_domains_from_state(curr_state)
+
+        likely_domains = [
+            domain for domain, depth in curr_domains
+        ]
+
+        @defer.inlineCallbacks
+        def try_backfill(domains):
+            # TODO: Should we try multiple of these at a time?
+            for dom in domains:
+                try:
+                    events = yield self.backfill(
+                        dom, room_id,
+                        limit=100,
+                        extremities=[e for e in extremities.keys()]
+                    )
+                except SynapseError:
+                    logger.info(
+                        "Failed to backfill from %s because %s",
+                        dom, e,
+                    )
+                    continue
+                except CodeMessageException as e:
+                    if 400 <= e.code < 500:
+                        raise
+
+                    logger.info(
+                        "Failed to backfill from %s because %s",
+                        dom, e,
+                    )
+                    continue
+                except NotRetryingDestination as e:
+                    logger.info(e.message)
+                    continue
+                except Exception as e:
+                    logger.warn(
+                        "Failed to backfill from %s because %s",
+                        dom, e,
+                    )
+                    continue
+
+                if events:
+                    defer.returnValue(True)
+            defer.returnValue(False)
+
+        success = yield try_backfill(likely_domains)
+        if success:
+            defer.returnValue(True)
+
+        # Huh, well *those* domains didn't work out. Lets try some domains
+        # from the time.
+
+        tried_domains = set(likely_domains)
+
+        event_ids = list(extremities.keys())
+
+        states = yield defer.gatherResults([
+            self.state_handler.resolve_state_groups([e])
+            for e in event_ids
+        ])
+        states = dict(zip(event_ids, [s[1] for s in states]))
+
+        for e_id, _ in sorted_extremeties_tuple:
+            likely_domains = get_domains_from_state(states[e_id])
+
+            success = yield try_backfill([
+                dom for dom in likely_domains
+                if dom not in tried_domains
+            ])
+            if success:
+                defer.returnValue(True)
+
+            tried_domains.update(likely_domains)
+
+        defer.returnValue(False)
+
+    @defer.inlineCallbacks
     def send_invite(self, target_host, event):
         """ Sends the invite to the remote server for signing.
 
@@ -431,9 +569,10 @@ class FederationHandler(BaseHandler):
                 auth_events=auth_events,
             )
 
-            d = self.notifier.on_new_room_event(
-                new_event, extra_users=[joinee]
-            )
+            with PreserveLoggingContext():
+                d = self.notifier.on_new_room_event(
+                    new_event, extra_users=[joinee]
+                )
 
             def log_failure(f):
                 logger.warn(
@@ -512,9 +651,10 @@ class FederationHandler(BaseHandler):
             target_user = UserID.from_string(target_user_id)
             extra_users.append(target_user)
 
-        d = self.notifier.on_new_room_event(
-            event, extra_users=extra_users
-        )
+        with PreserveLoggingContext():
+            d = self.notifier.on_new_room_event(
+                event, extra_users=extra_users
+            )
 
         def log_failure(f):
             logger.warn(
@@ -594,9 +734,10 @@ class FederationHandler(BaseHandler):
         )
 
         target_user = UserID.from_string(event.state_key)
-        d = self.notifier.on_new_room_event(
-            event, extra_users=[target_user],
-        )
+        with PreserveLoggingContext():
+            d = self.notifier.on_new_room_event(
+                event, extra_users=[target_user],
+            )
 
         def log_failure(f):
             logger.warn(
@@ -921,7 +1062,7 @@ class FederationHandler(BaseHandler):
                     if d in have_events and not have_events[d]
                 ],
                 consumeErrors=True
-            )
+            ).addErrback(unwrapFirstError)
 
             if different_events:
                 local_view = dict(auth_events)