Merge pull request #2115 from matrix-org/erikj/dedupe_federation_repl

Reduce federation replication traffic

synapse/app/federation_sender.py

@@ -28,6 +28,7 @@ from synapse.replication.slave.storage.deviceinbox import SlavedDeviceInboxStore
 from synapse.replication.slave.storage.events import SlavedEventStore
 from synapse.replication.slave.storage.receipts import SlavedReceiptsStore
 from synapse.replication.slave.storage.registration import SlavedRegistrationStore
+from synapse.replication.slave.storage.presence import SlavedPresenceStore
 from synapse.replication.slave.storage.transactions import TransactionStore
 from synapse.replication.slave.storage.devices import SlavedDeviceStore
 from synapse.replication.tcp.client import ReplicationClientHandler
@@ -55,7 +56,7 @@ logger = logging.getLogger("synapse.app.appservice")
 
 class FederationSenderSlaveStore(
     SlavedDeviceInboxStore, TransactionStore, SlavedReceiptsStore, SlavedEventStore,
-    SlavedRegistrationStore, SlavedDeviceStore,
+    SlavedRegistrationStore, SlavedDeviceStore, SlavedPresenceStore,
 ):
     def __init__(self, db_conn, hs):
         super(FederationSenderSlaveStore, self).__init__(db_conn, hs)

synapse/app/synchrotron.py

@@ -20,7 +20,7 @@ from synapse.api.constants import EventTypes
 from synapse.config._base import ConfigError
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.logger import setup_logging
-from synapse.handlers.presence import PresenceHandler
+from synapse.handlers.presence import PresenceHandler, get_interested_parties
 from synapse.http.site import SynapseSite
 from synapse.http.server import JsonResource
 from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
@@ -44,7 +44,7 @@ from synapse.replication.tcp.client import ReplicationClientHandler
 from synapse.server import HomeServer
 from synapse.storage.client_ips import ClientIpStore
 from synapse.storage.engines import create_engine
-from synapse.storage.presence import PresenceStore, UserPresenceState
+from synapse.storage.presence import UserPresenceState
 from synapse.storage.roommember import RoomMemberStore
 from synapse.util.httpresourcetree import create_resource_tree
 from synapse.util.logcontext import LoggingContext, PreserveLoggingContext, preserve_fn
@@ -89,16 +89,6 @@ class SynchrotronSlavedStore(
         RoomMemberStore.__dict__["did_forget"]
     )
 
-    # XXX: This is a bit broken because we don't persist the accepted list in a
-    # way that can be replicated. This means that we don't have a way to
-    # invalidate the cache correctly.
-    get_presence_list_accepted = PresenceStore.__dict__[
-        "get_presence_list_accepted"
-    ]
-    get_presence_list_observers_accepted = PresenceStore.__dict__[
-        "get_presence_list_observers_accepted"
-    ]
-
 
 UPDATE_SYNCING_USERS_MS = 10 * 1000
 
@@ -172,7 +162,6 @@ class SynchrotronPresence(object):
 
     get_states = PresenceHandler.get_states.__func__
     get_state = PresenceHandler.get_state.__func__
-    _get_interested_parties = PresenceHandler._get_interested_parties.__func__
     current_state_for_users = PresenceHandler.current_state_for_users.__func__
 
     def user_syncing(self, user_id, affect_presence):
@@ -206,10 +195,8 @@ class SynchrotronPresence(object):
 
     @defer.inlineCallbacks
     def notify_from_replication(self, states, stream_id):
-        parties = yield self._get_interested_parties(
+        parties = yield get_interested_parties(self.store, states)
-            states, calculate_remote_hosts=False
+        room_ids_to_states, users_to_states = parties
-        )
-        room_ids_to_states, users_to_states, _ = parties
 
         self.notifier.on_new_event(
             "presence_key", stream_id, rooms=room_ids_to_states.keys(),

synapse/federation/send_queue.py

@@ -53,18 +53,19 @@ class FederationRemoteSendQueue(object):
         self.server_name = hs.hostname
         self.clock = hs.get_clock()
         self.notifier = hs.get_notifier()
+        self.is_mine_id = hs.is_mine_id
 
-        self.presence_map = {}
+        self.presence_map = {}  # Pending presence map user_id -> UserPresenceState
-        self.presence_changed = sorteddict()
+        self.presence_changed = sorteddict()  # Stream position -> user_id
 
-        self.keyed_edu = {}
+        self.keyed_edu = {}  # (destination, key) -> EDU
-        self.keyed_edu_changed = sorteddict()
+        self.keyed_edu_changed = sorteddict()  # stream position -> (destination, key)
 
-        self.edus = sorteddict()
+        self.edus = sorteddict()  # stream position -> Edu
 
-        self.failures = sorteddict()
+        self.failures = sorteddict()  # stream position -> (destination, Failure)
 
-        self.device_messages = sorteddict()
+        self.device_messages = sorteddict()  # stream position -> destination
 
         self.pos = 1
         self.pos_time = sorteddict()
@@ -120,7 +121,9 @@ class FederationRemoteSendQueue(object):
                 del self.presence_changed[key]
 
             user_ids = set(
-                user_id for uids in self.presence_changed.values() for _, user_id in uids
+                user_id
+                for uids in self.presence_changed.itervalues()
+                for user_id in uids
             )
 
             to_del = [
@@ -187,18 +190,20 @@ class FederationRemoteSendQueue(object):
 
         self.notifier.on_new_replication_data()
 
-    def send_presence(self, destination, states):
+    def send_presence(self, states):
-        """As per TransactionQueue"""
+        """As per TransactionQueue
+
+        Args:
+            states (list(UserPresenceState))
+        """
         pos = self._next_pos()
 
-        self.presence_map.update({
+        # We only want to send presence for our own users, so lets always just
-            state.user_id: state
+        # filter here just in case.
-            for state in states
+        local_states = filter(lambda s: self.is_mine_id(s.user_id), states)
-        })
 
-        self.presence_changed[pos] = [
+        self.presence_map.update({state.user_id: state for state in local_states})
-            (destination, state.user_id) for state in states
+        self.presence_changed[pos] = [state.user_id for state in local_states]
-        ]
 
         self.notifier.on_new_replication_data()
 
@@ -251,15 +256,14 @@ class FederationRemoteSendQueue(object):
         keys = self.presence_changed.keys()
         i = keys.bisect_right(from_token)
         j = keys.bisect_right(to_token) + 1
-        dest_user_ids = set(
+        dest_user_ids = [
-            (pos, dest_user_id)
+            (pos, user_id)
             for pos in keys[i:j]
-            for dest_user_id in self.presence_changed[pos]
+            for user_id in self.presence_changed[pos]
-        )
+        ]
 
-        for (key, (dest, user_id)) in dest_user_ids:
+        for (key, user_id) in dest_user_ids:
             rows.append((key, PresenceRow(
-                destination=dest,
                 state=self.presence_map[user_id],
             )))
 
@@ -357,7 +361,6 @@ class BaseFederationRow(object):
 
 
 class PresenceRow(BaseFederationRow, namedtuple("PresenceRow", (
-    "destination",  # str
     "state",  # UserPresenceState
 ))):
     TypeId = "p"
@@ -365,18 +368,14 @@ class PresenceRow(BaseFederationRow, namedtuple("PresenceRow", (
     @staticmethod
     def from_data(data):
         return PresenceRow(
-            destination=data["destination"],
+            state=UserPresenceState.from_dict(data)
-            state=UserPresenceState.from_dict(data["state"])
         )
 
     def to_data(self):
-        return {
+        return self.state.as_dict()
-            "destination": self.destination,
-            "state": self.state.as_dict()
-        }
 
     def add_to_buffer(self, buff):
-        buff.presence.setdefault(self.destination, []).append(self.state)
+        buff.presence.append(self.state)
 
 
 class KeyedEduRow(BaseFederationRow, namedtuple("KeyedEduRow", (
@@ -487,7 +486,7 @@ TypeToRow = {
 
 
 ParsedFederationStreamData = namedtuple("ParsedFederationStreamData", (
-    "presence",  # dict of destination -> [UserPresenceState]
+    "presence",  # list(UserPresenceState)
     "keyed_edus",  # dict of destination -> { key -> Edu }
     "edus",  # dict of destination -> [Edu]
     "failures",  # dict of destination -> [failures]
@@ -509,7 +508,7 @@ def process_rows_for_federation(transaction_queue, rows):
     # them into the appropriate collection and then send them off.
 
     buff = ParsedFederationStreamData(
-        presence={},
+        presence=[],
         keyed_edus={},
         edus={},
         failures={},
@@ -526,8 +525,8 @@ def process_rows_for_federation(transaction_queue, rows):
         parsed_row = RowType.from_data(row.data)
         parsed_row.add_to_buffer(buff)
 
-    for destination, states in buff.presence.iteritems():
+    if buff.presence:
-        transaction_queue.send_presence(destination, states)
+        transaction_queue.send_presence(buff.presence)
 
     for destination, edu_map in buff.keyed_edus.iteritems():
         for key, edu in edu_map.items():

synapse/federation/transaction_queue.py

@@ -21,11 +21,11 @@ from .units import Transaction, Edu
 
 from synapse.api.errors import HttpResponseException
 from synapse.util.async import run_on_reactor
-from synapse.util.logcontext import preserve_context_over_fn
+from synapse.util.logcontext import preserve_context_over_fn, preserve_fn
 from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
 from synapse.util.metrics import measure_func
 from synapse.types import get_domain_from_id
-from synapse.handlers.presence import format_user_presence_state
+from synapse.handlers.presence import format_user_presence_state, get_interested_remotes
 import synapse.metrics
 
 import logging
@@ -79,8 +79,18 @@ class TransactionQueue(object):
         # destination -> list of tuple(edu, deferred)
         self.pending_edus_by_dest = edus = {}
 
-        # Presence needs to be separate as we send single aggragate EDUs
+        # Map of user_id -> UserPresenceState for all the pending presence
+        # to be sent out by user_id. Entries here get processed and put in
+        # pending_presence_by_dest
+        self.pending_presence = {}
+
+        # Map of destination -> user_id -> UserPresenceState of pending presence
+        # to be sent to each destinations
         self.pending_presence_by_dest = presence = {}
+
+        # Pending EDUs by their "key". Keyed EDUs are EDUs that get clobbered
+        # based on their key (e.g. typing events by room_id)
+        # Map of destination -> (edu_type, key) -> Edu
         self.pending_edus_keyed_by_dest = edus_keyed = {}
 
         metrics.register_callback(
@@ -115,6 +125,8 @@ class TransactionQueue(object):
         self._is_processing = False
         self._last_poked_id = -1
 
+        self._processing_pending_presence = False
+
     def can_send_to(self, destination):
         """Can we send messages to the given server?
 
@@ -226,17 +238,71 @@ class TransactionQueue(object):
                 self._attempt_new_transaction, destination
             )
 
-    def send_presence(self, destination, states):
+    @preserve_fn  # the caller should not yield on this
-        if not self.can_send_to(destination):
+    @defer.inlineCallbacks
+    def send_presence(self, states):
+        """Send the new presence states to the appropriate destinations.
+
+        This actually queues up the presence states ready for sending and
+        triggers a background task to process them and send out the transactions.
+
+        Args:
+            states (list(UserPresenceState))
+        """
+
+        # First we queue up the new presence by user ID, so multiple presence
+        # updates in quick successtion are correctly handled
+        # We only want to send presence for our own users, so lets always just
+        # filter here just in case.
+        self.pending_presence.update({
+            state.user_id: state for state in states
+            if self.is_mine_id(state.user_id)
+        })
+
+        # We then handle the new pending presence in batches, first figuring
+        # out the destinations we need to send each state to and then poking it
+        # to attempt a new transaction. We linearize this so that we don't
+        # accidentally mess up the ordering and send multiple presence updates
+        # in the wrong order
+        if self._processing_pending_presence:
             return
 
-        self.pending_presence_by_dest.setdefault(destination, {}).update({
+        self._processing_pending_presence = True
+        try:
+            while True:
+                states_map = self.pending_presence
+                self.pending_presence = {}
+
+                if not states_map:
+                    break
+
+                yield self._process_presence_inner(states_map.values())
+        finally:
+            self._processing_pending_presence = False
+
+    @measure_func("txnqueue._process_presence")
+    @defer.inlineCallbacks
+    def _process_presence_inner(self, states):
+        """Given a list of states populate self.pending_presence_by_dest and
+        poke to send a new transaction to each destination
+
+        Args:
+            states (list(UserPresenceState))
+        """
+        hosts_and_states = yield get_interested_remotes(self.store, states)
+
+        for destinations, states in hosts_and_states:
+            for destination in destinations:
+                if not self.can_send_to(destination):
+                    continue
+
+                self.pending_presence_by_dest.setdefault(
+                    destination, {}
+                ).update({
                     state.user_id: state for state in states
                 })
 
-        preserve_context_over_fn(
+                preserve_fn(self._attempt_new_transaction)(destination)
-            self._attempt_new_transaction, destination
-        )
 
     def send_edu(self, destination, edu_type, content, key=None):
         edu = Edu(

synapse/handlers/presence.py

@@ -318,11 +318,7 @@ class PresenceHandler(object):
             if to_federation_ping:
                 federation_presence_out_counter.inc_by(len(to_federation_ping))
 
-                _, _, hosts_to_states = yield self._get_interested_parties(
+                self._push_to_remotes(to_federation_ping.values())
-                    to_federation_ping.values()
-                )
-
-                self._push_to_remotes(hosts_to_states)
 
     def _handle_timeouts(self):
         """Checks the presence of users that have timed out and updates as
@@ -614,54 +610,6 @@ class PresenceHandler(object):
 
         defer.returnValue(states)
 
-    @defer.inlineCallbacks
-    def _get_interested_parties(self, states, calculate_remote_hosts=True):
-        """Given a list of states return which entities (rooms, users, servers)
-        are interested in the given states.
-
-        Returns:
-            3-tuple: `(room_ids_to_states, users_to_states, hosts_to_states)`,
-            with each item being a dict of `entity_name` -> `[UserPresenceState]`
-        """
-        room_ids_to_states = {}
-        users_to_states = {}
-        for state in states:
-            room_ids = yield self.store.get_rooms_for_user(state.user_id)
-            for room_id in room_ids:
-                room_ids_to_states.setdefault(room_id, []).append(state)
-
-            plist = yield self.store.get_presence_list_observers_accepted(state.user_id)
-            for u in plist:
-                users_to_states.setdefault(u, []).append(state)
-
-            # Always notify self
-            users_to_states.setdefault(state.user_id, []).append(state)
-
-        hosts_to_states = {}
-        if calculate_remote_hosts:
-            for room_id, states in room_ids_to_states.items():
-                local_states = filter(lambda s: self.is_mine_id(s.user_id), states)
-                if not local_states:
-                    continue
-
-                hosts = yield self.store.get_hosts_in_room(room_id)
-
-                for host in hosts:
-                    hosts_to_states.setdefault(host, []).extend(local_states)
-
-        for user_id, states in users_to_states.items():
-            local_states = filter(lambda s: self.is_mine_id(s.user_id), states)
-            if not local_states:
-                continue
-
-            host = get_domain_from_id(user_id)
-            hosts_to_states.setdefault(host, []).extend(local_states)
-
-        # TODO: de-dup hosts_to_states, as a single host might have multiple
-        # of same presence
-
-        defer.returnValue((room_ids_to_states, users_to_states, hosts_to_states))
-
     @defer.inlineCallbacks
     def _persist_and_notify(self, states):
         """Persist states in the database, poke the notifier and send to
@@ -669,34 +617,33 @@ class PresenceHandler(object):
         """
         stream_id, max_token = yield self.store.update_presence(states)
 
-        parties = yield self._get_interested_parties(states)
+        parties = yield get_interested_parties(self.store, states)
-        room_ids_to_states, users_to_states, hosts_to_states = parties
+        room_ids_to_states, users_to_states = parties
 
         self.notifier.on_new_event(
             "presence_key", stream_id, rooms=room_ids_to_states.keys(),
-            users=[UserID.from_string(u) for u in users_to_states.keys()]
+            users=[UserID.from_string(u) for u in users_to_states]
         )
 
-        self._push_to_remotes(hosts_to_states)
+        self._push_to_remotes(states)
 
     @defer.inlineCallbacks
     def notify_for_states(self, state, stream_id):
-        parties = yield self._get_interested_parties([state])
+        parties = yield get_interested_parties(self.store, [state])
-        room_ids_to_states, users_to_states, hosts_to_states = parties
+        room_ids_to_states, users_to_states = parties
 
         self.notifier.on_new_event(
             "presence_key", stream_id, rooms=room_ids_to_states.keys(),
-            users=[UserID.from_string(u) for u in users_to_states.keys()]
+            users=[UserID.from_string(u) for u in users_to_states]
         )
 
-    def _push_to_remotes(self, hosts_to_states):
+    def _push_to_remotes(self, states):
         """Sends state updates to remote servers.
 
         Args:
-            hosts_to_states (dict): Mapping `server_name` -> `[UserPresenceState]`
+            states (list(UserPresenceState))
         """
-        for host, states in hosts_to_states.items():
+        self.federation.send_presence(states)
-            self.federation.send_presence(host, states)
 
     @defer.inlineCallbacks
     def incoming_presence(self, origin, content):
@@ -837,14 +784,13 @@ class PresenceHandler(object):
         if self.is_mine(user):
             state = yield self.current_state_for_user(user.to_string())
 
-            hosts = set(get_domain_from_id(u) for u in user_ids)
+            self._push_to_remotes([state])
-            self._push_to_remotes({host: (state,) for host in hosts})
         else:
             user_ids = filter(self.is_mine_id, user_ids)
 
             states = yield self.current_state_for_users(user_ids)
 
-            self._push_to_remotes({user.domain: states.values()})
+            self._push_to_remotes(states.values())
 
     @defer.inlineCallbacks
     def get_presence_list(self, observer_user, accepted=None):
@@ -1344,3 +1290,66 @@ def handle_update(prev_state, new_state, is_mine, wheel_timer, now):
         persist_and_notify = True
 
     return new_state, persist_and_notify, federation_ping
+
+
+@defer.inlineCallbacks
+def get_interested_parties(store, states):
+    """Given a list of states return which entities (rooms, users)
+    are interested in the given states.
+
+    Args:
+        states (list(UserPresenceState))
+
+    Returns:
+        2-tuple: `(room_ids_to_states, users_to_states)`,
+        with each item being a dict of `entity_name` -> `[UserPresenceState]`
+    """
+    room_ids_to_states = {}
+    users_to_states = {}
+    for state in states:
+        room_ids = yield store.get_rooms_for_user(state.user_id)
+        for room_id in room_ids:
+            room_ids_to_states.setdefault(room_id, []).append(state)
+
+        plist = yield store.get_presence_list_observers_accepted(state.user_id)
+        for u in plist:
+            users_to_states.setdefault(u, []).append(state)
+
+        # Always notify self
+        users_to_states.setdefault(state.user_id, []).append(state)
+
+    defer.returnValue((room_ids_to_states, users_to_states))
+
+
+@defer.inlineCallbacks
+def get_interested_remotes(store, states):
+    """Given a list of presence states figure out which remote servers
+    should be sent which.
+
+    All the presence states should be for local users only.
+
+    Args:
+        store (DataStore)
+        states (list(UserPresenceState))
+
+    Returns:
+        Deferred list of ([destinations], [UserPresenceState]), where for
+        each row the list of UserPresenceState should be sent to each
+        destination
+    """
+    hosts_and_states = []
+
+    # First we look up the rooms each user is in (as well as any explicit
+    # subscriptions), then for each distinct room we look up the remote
+    # hosts in those rooms.
+    room_ids_to_states, users_to_states = yield get_interested_parties(store, states)
+
+    for room_id, states in room_ids_to_states.iteritems():
+        hosts = yield store.get_hosts_in_room(room_id)
+        hosts_and_states.append((hosts, states))
+
+    for user_id, states in users_to_states.iteritems():
+        host = get_domain_from_id(user_id)
+        hosts_and_states.append(([host], states))
+
+    defer.returnValue(hosts_and_states)

synapse/replication/slave/storage/events.py

@@ -71,6 +71,7 @@ class SlavedEventStore(BaseSlavedStore):
     # to reach inside the __dict__ to extract them.
     get_rooms_for_user = RoomMemberStore.__dict__["get_rooms_for_user"]
     get_users_in_room = RoomMemberStore.__dict__["get_users_in_room"]
+    get_hosts_in_room = RoomMemberStore.__dict__["get_hosts_in_room"]
     get_users_who_share_room_with_user = (
         RoomMemberStore.__dict__["get_users_who_share_room_with_user"]
     )

synapse/replication/slave/storage/presence.py

@@ -39,6 +39,16 @@ class SlavedPresenceStore(BaseSlavedStore):
     _get_presence_for_user = PresenceStore.__dict__["_get_presence_for_user"]
     get_presence_for_users = PresenceStore.__dict__["get_presence_for_users"]
 
+    # XXX: This is a bit broken because we don't persist the accepted list in a
+    # way that can be replicated. This means that we don't have a way to
+    # invalidate the cache correctly.
+    get_presence_list_accepted = PresenceStore.__dict__[
+        "get_presence_list_accepted"
+    ]
+    get_presence_list_observers_accepted = PresenceStore.__dict__[
+        "get_presence_list_observers_accepted"
+    ]
+
     def get_current_presence_token(self):
         return self._presence_id_gen.get_current_token()