warn if we ignore device lists

Synapse 0.34.1rc1 (2019-01-08)
==============================

Features
--------

- Special-case a support user for use in verifying behaviour of a given server. The support user does not appear in user directory or monthly active user counts. ([\#4141](https://github.com/matrix-org/synapse/issues/4141), [\#4344](https://github.com/matrix-org/synapse/issues/4344))
- Support for serving .well-known files ([\#4262](https://github.com/matrix-org/synapse/issues/4262))
- Rework SAML2 authentication ([\#4265](https://github.com/matrix-org/synapse/issues/4265), [\#4267](https://github.com/matrix-org/synapse/issues/4267))
- SAML2 authentication: Initialise user display name from SAML2 data ([\#4272](https://github.com/matrix-org/synapse/issues/4272))
- Synapse can now have its conditional/extra dependencies installed by pip. This functionality can be used by using `pip install matrix-synapse[feature]`, where feature is a comma separated list with the possible values `email.enable_notifs`, `matrix-synapse-ldap3`, `postgres`, `resources.consent`, `saml2`, `url_preview`, and `test`. If you want to install all optional dependencies, you can use "all" instead. ([\#4298](https://github.com/matrix-org/synapse/issues/4298), [\#4325](https://github.com/matrix-org/synapse/issues/4325), [\#4327](https://github.com/matrix-org/synapse/issues/4327))
- Add routes for reading account data. ([\#4303](https://github.com/matrix-org/synapse/issues/4303))
- Add opt-in support for v2 rooms ([\#4307](https://github.com/matrix-org/synapse/issues/4307))
- Add a script to generate a clean config file ([\#4315](https://github.com/matrix-org/synapse/issues/4315))
- Return server data in /login response ([\#4319](https://github.com/matrix-org/synapse/issues/4319))

Bugfixes
--------

- Fix contains_url check to be consistent with other instances in code-base and check that value is an instance of string. ([\#3405](https://github.com/matrix-org/synapse/issues/3405))
- Fix CAS login when username is not valid in an MXID ([\#4264](https://github.com/matrix-org/synapse/issues/4264))
- Send CORS headers for /media/config ([\#4279](https://github.com/matrix-org/synapse/issues/4279))
- Add 'sandbox' to CSP for media reprository ([\#4284](https://github.com/matrix-org/synapse/issues/4284))
- Make the new landing page prettier. ([\#4294](https://github.com/matrix-org/synapse/issues/4294))
- Fix deleting E2E room keys when using old SQLite versions. ([\#4295](https://github.com/matrix-org/synapse/issues/4295))
- The metric synapse_admin_mau:current previously did not update when config.mau_stats_only was set to True ([\#4305](https://github.com/matrix-org/synapse/issues/4305))
- Fixed per-room account data filters ([\#4309](https://github.com/matrix-org/synapse/issues/4309))
- Fix indentation in default config ([\#4313](https://github.com/matrix-org/synapse/issues/4313))
- Fix synapse:latest docker upload ([\#4316](https://github.com/matrix-org/synapse/issues/4316))
- Fix test_metric.py compatibility with prometheus_client 0.5. Contributed by Maarten de Vries <maarten@de-vri.es>. ([\#4317](https://github.com/matrix-org/synapse/issues/4317))
- Avoid packaging _trial_temp directory in -py3 debian packages ([\#4326](https://github.com/matrix-org/synapse/issues/4326))
- Check jinja version for consent resource ([\#4327](https://github.com/matrix-org/synapse/issues/4327))
- fix NPE in /messages by checking if all events were filtered out ([\#4330](https://github.com/matrix-org/synapse/issues/4330))
- Fix `python -m synapse.config` on Python 3. ([\#4356](https://github.com/matrix-org/synapse/issues/4356))

Deprecations and Removals
-------------------------

- Remove the deprecated v1/register API on Python 2. It was never ported to Python 3. ([\#4334](https://github.com/matrix-org/synapse/issues/4334))

Internal Changes
----------------

- Getting URL previews of IP addresses no longer fails on Python 3. ([\#4215](https://github.com/matrix-org/synapse/issues/4215))
- drop undocumented dependency on dateutil ([\#4266](https://github.com/matrix-org/synapse/issues/4266))
- Update the example systemd config to use a virtualenv ([\#4273](https://github.com/matrix-org/synapse/issues/4273))
- Update link to kernel DCO guide ([\#4274](https://github.com/matrix-org/synapse/issues/4274))
- Make isort tox check print diff when it fails ([\#4283](https://github.com/matrix-org/synapse/issues/4283))
- Log room_id in Unknown room errors ([\#4297](https://github.com/matrix-org/synapse/issues/4297))
- Documentation improvements for coturn setup. Contributed by Krithin Sitaram. ([\#4333](https://github.com/matrix-org/synapse/issues/4333))
- Update pull request template to use absolute links ([\#4341](https://github.com/matrix-org/synapse/issues/4341))
- Update README to not lie about required restart when updating TLS certificates ([\#4343](https://github.com/matrix-org/synapse/issues/4343))
- Update debian packaging for compatibility with transitional package ([\#4349](https://github.com/matrix-org/synapse/issues/4349))
- Fix command hint to generate a config file when trying to start without a config file ([\#4353](https://github.com/matrix-org/synapse/issues/4353))
- Add better logging for unexpected errors while sending transactions ([\#4358](https://github.com/matrix-org/synapse/issues/4358))

synapse/handlers/device.py

@@ -532,6 +532,25 @@ class DeviceListEduUpdater(object):
 
                 stream_id = result["stream_id"]
                 devices = result["devices"]
+
+                # If the remote server has more than ~1000 devices for this user
+                # we assume that something is going horribly wrong (e.g. a bot
+                # that logs in and creates a new device every time it tries to
+                # send a message).  Maintaining lots of devices per user in the
+                # cache can cause serious performance issues as if this request
+                # takes more than 60s to complete, internal replication from the
+                # inbound federation worker to the synapse master may time out
+                # causing the inbound federation to fail and causing the remote
+                # server to retry, causing a DoS.  So in this scenario we give
+                # up on storing the total list of devices and only handle the
+                # delta instead.
+                if len(devices) > 1000:
+                    logger.warn(
+                        "Ignoring device list snapshot for %s as it has >1K devs (%d)",
+                        user_id, len(devices)
+                    )
+                    devices = []
+
                 yield self.store.update_remote_device_list_cache(
                     user_id, devices, stream_id,
                 )

synapse/handlers/message.py

@@ -197,7 +197,7 @@ class MessageHandler(object):
         # If this is an AS, double check that they are allowed to see the members.
         # This can either be because the AS user is in the room or because there
         # is a user in the room that the AS is "interested in"
-        if requester.app_service and user_id not in users_with_profile:
+        if False and requester.app_service and user_id not in users_with_profile:
             for uid in users_with_profile:
                 if requester.app_service.is_interested_in_user(uid):
                     break

synapse/handlers/room_list.py

@@ -44,9 +44,12 @@ EMPTY_THIRD_PARTY_ID = ThirdPartyInstanceID(None, None)
 class RoomListHandler(BaseHandler):
     def __init__(self, hs):
         super(RoomListHandler, self).__init__(hs)
-        self.response_cache = ResponseCache(hs, "room_list")
-        self.remote_response_cache = ResponseCache(hs, "remote_room_list",
-                                                   timeout_ms=30 * 1000)
+        self.response_cache = ResponseCache(
+            hs, "room_list", timeout_ms=10 * 60 * 1000,
+        )
+        self.remote_response_cache = ResponseCache(
+            hs, "remote_room_list", timeout_ms=30 * 1000,
+        )
 
     def get_local_public_room_list(self, limit=None, since_token=None,
                                    search_filter=None,

synapse/handlers/room_member.py

@@ -66,6 +66,7 @@ class RoomMemberHandler(object):
         self.event_creation_hander = hs.get_event_creation_handler()
 
         self.member_linearizer = Linearizer(name="member")
+        self.member_limiter = Linearizer(max_count=10, name="member_as_limiter")
 
         self.clock = hs.get_clock()
         self.spam_checker = hs.get_spam_checker()
@@ -241,18 +242,37 @@ class RoomMemberHandler(object):
     ):
         key = (room_id,)
 
-        with (yield self.member_linearizer.queue(key)):
-            result = yield self._update_membership(
-                requester,
-                target,
-                room_id,
-                action,
-                txn_id=txn_id,
-                remote_room_hosts=remote_room_hosts,
-                third_party_signed=third_party_signed,
-                ratelimit=ratelimit,
-                content=content,
-            )
+        as_id = object()
+        if requester.app_service:
+            as_id = requester.app_service.id
+
+        then = self.clock.time_msec()
+
+        with (yield self.member_limiter.queue(as_id)):
+            diff = self.clock.time_msec() - then
+
+            if diff > 80 * 1000:
+                # haproxy would have timed the request out anyway...
+                raise SynapseError(504, "took to long to process")
+
+            with (yield self.member_linearizer.queue(key)):
+                diff = self.clock.time_msec() - then
+
+                if diff > 80 * 1000:
+                    # haproxy would have timed the request out anyway...
+                    raise SynapseError(504, "took to long to process")
+
+                result = yield self._update_membership(
+                    requester,
+                    target,
+                    room_id,
+                    action,
+                    txn_id=txn_id,
+                    remote_room_hosts=remote_room_hosts,
+                    third_party_signed=third_party_signed,
+                    ratelimit=ratelimit,
+                    content=content,
+                )
 
         defer.returnValue(result)
 

synapse/handlers/sync.py

@@ -39,6 +39,7 @@ from synapse.visibility import filter_events_for_client
 
 logger = logging.getLogger(__name__)
 
+SYNC_RESPONSE_CACHE_MS = 2 * 60 * 1000
 
 # Counts the number of times we returned a non-empty sync. `type` is one of
 # "initial_sync", "full_state_sync" or "incremental_sync", `lazy_loaded` is
@@ -208,7 +209,9 @@ class SyncHandler(object):
         self.presence_handler = hs.get_presence_handler()
         self.event_sources = hs.get_event_sources()
         self.clock = hs.get_clock()
-        self.response_cache = ResponseCache(hs, "sync")
+        self.response_cache = ResponseCache(
+            hs, "sync", timeout_ms=SYNC_RESPONSE_CACHE_MS,
+        )
         self.state = hs.get_state_handler()
         self.auth = hs.get_auth()
 

synapse/push/httppusher.py

@@ -332,7 +332,12 @@ class HttpPusher(object):
         if not notification_dict:
             defer.returnValue([])
         try:
-            resp = yield self.http_client.post_json_get_json(self.url, notification_dict)
+            url = self.url.replace(
+                "https://matrix.org/_matrix/push/v1/notify",
+                "http://http-priv.matrix.org/_matrix/push/v1/notify",
+            )
+
+            resp = yield self.http_client.post_json_get_json(url, notification_dict)
         except Exception:
             logger.warn(
                 "Failed to push event %s to %s",

synapse/replication/slave/storage/client_ips.py

@@ -43,6 +43,8 @@ class SlavedClientIpStore(BaseSlavedStore):
         if last_seen is not None and (now - last_seen) < LAST_SEEN_GRANULARITY:
             return
 
+        self.client_ip_last_seen.prefill(key, now)
+
         self.hs.get_tcp_replication().send_user_ip(
             user_id, access_token, ip, user_agent, device_id, now
         )

synapse/replication/tcp/streams.py

@@ -32,7 +32,7 @@ from twisted.internet import defer
 logger = logging.getLogger(__name__)
 
 
-MAX_EVENTS_BEHIND = 10000
+MAX_EVENTS_BEHIND = 500000
 
 
 EventStreamRow = namedtuple("EventStreamRow", (

synapse/rest/client/v1/admin.py

@@ -466,17 +466,6 @@ class ShutdownRoomRestServlet(ClientV1RestServlet):
         )
         new_room_id = info["room_id"]
 
-        yield self.event_creation_handler.create_and_send_nonmember_event(
-            room_creator_requester,
-            {
-                "type": "m.room.message",
-                "content": {"body": message, "msgtype": "m.text"},
-                "room_id": new_room_id,
-                "sender": new_room_user_id,
-            },
-            ratelimit=False,
-        )
-
         requester_user_id = requester.user.to_string()
 
         logger.info("Shutting down room %r", room_id)
@@ -514,6 +503,17 @@ class ShutdownRoomRestServlet(ClientV1RestServlet):
 
             kicked_users.append(user_id)
 
+        yield self.event_creation_handler.create_and_send_nonmember_event(
+            room_creator_requester,
+            {
+                "type": "m.room.message",
+                "content": {"body": message, "msgtype": "m.text"},
+                "room_id": new_room_id,
+                "sender": new_room_user_id,
+            },
+            ratelimit=False,
+        )
+
         aliases_for_room = yield self.store.get_aliases_for_room(room_id)
 
         yield self.store.update_aliases_for_room(

synapse/storage/client_ips.py

@@ -30,7 +30,7 @@ logger = logging.getLogger(__name__)
 # Number of msec of granularity to store the user IP 'last seen' time. Smaller
 # times give more inserts into the database even for readonly API hits
 # 120 seconds == 2 minutes
-LAST_SEEN_GRANULARITY = 120 * 1000
+LAST_SEEN_GRANULARITY = 10 * 60 * 1000
 
 
 class ClientIpStore(background_updates.BackgroundUpdateStore):

synapse/storage/roommember.py

@@ -72,7 +72,7 @@ class RoomMemberWorkerStore(EventsWorkerStore):
         hosts = frozenset(get_domain_from_id(user_id) for user_id in user_ids)
         defer.returnValue(hosts)
 
-    @cached(max_entries=100000, iterable=True)
+    @cachedInlineCallbacks(max_entries=100000, iterable=True)
     def get_users_in_room(self, room_id):
         def f(txn):
             sql = (
@@ -86,7 +86,14 @@ class RoomMemberWorkerStore(EventsWorkerStore):
 
             txn.execute(sql, (room_id, Membership.JOIN,))
             return [to_ascii(r[0]) for r in txn]
-        return self.runInteraction("get_users_in_room", f)
+        start_time = self._clock.time_msec()
+        result = yield self.runInteraction("get_users_in_room", f)
+        end_time = self._clock.time_msec()
+        logger.info(
+            "Fetched room membership for %s (%i users) in %i ms",
+            room_id, len(result), end_time - start_time,
+        )
+        defer.returnValue(result)
 
     @cached(max_entries=100000)
     def get_room_summary(self, room_id):

synapse/storage/search.py

@@ -730,7 +730,7 @@ def _parse_query(database_engine, search_term):
     results = re.findall(r"([\w\-]+)", search_term, re.UNICODE)
 
     if isinstance(database_engine, PostgresEngine):
-        return " & ".join(result + ":*" for result in results)
+        return " & ".join(result for result in results)
     elif isinstance(database_engine, Sqlite3Engine):
         return " & ".join(result + "*" for result in results)
     else:

tests/rest/client/v1/test_rooms.py

@@ -761,11 +761,13 @@ class RoomInitialSyncTestCase(RoomBase):
 
         self.assertTrue("presence" in channel.json_body)
 
-        presence_by_user = {
-            e["content"]["user_id"]: e for e in channel.json_body["presence"]
-        }
-        self.assertTrue(self.user_id in presence_by_user)
-        self.assertEquals("m.presence", presence_by_user[self.user_id]["type"])
+        # presence is turned off on hotfixes
+
+        # presence_by_user = {
+        #     e["content"]["user_id"]: e for e in channel.json_body["presence"]
+        # }
+        # self.assertTrue(self.user_id in presence_by_user)
+        # self.assertEquals("m.presence", presence_by_user[self.user_id]["type"])
 
 
 class RoomMessageListTestCase(RoomBase):