diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py
index 4ebb0e8bc0..f7065517e5 100644
--- a/synapse/federation/sender/__init__.py
+++ b/synapse/federation/sender/__init__.py
@@ -152,9 +152,24 @@ class FederationSender(object):
@defer.inlineCallbacks
def _process_event_queue_loop(self):
+ loop_start_time = self.clock.time_msec()
try:
self._is_processing = True
while True:
+ # if we've been going around this loop for a long time without
+ # catching up, deprioritise transaction transmission. This should mean
+ # that events get batched into fewer transactions, which is more
+ # efficient, and hence give us a chance to catch up
+ if (
+ self.clock.time_msec() - loop_start_time > 60 * 1000
+ and not self._transaction_manager.deprioritise_transmission
+ ):
+ logger.warning(
+ "Event queue is getting behind: deprioritising transaction "
+ "transmission"
+ )
+ self._transaction_manager.deprioritise_transmission = True
+
last_token = yield self.store.get_federation_out_pos("events")
next_token, events = yield self.store.get_all_new_events_stream(
last_token, self._last_poked_id, limit=100
@@ -252,6 +267,9 @@ class FederationSender(object):
finally:
self._is_processing = False
+ if self._transaction_manager.deprioritise_transmission:
+ logger.info("Event queue caught up: re-prioritising transmission")
+ self._transaction_manager.deprioritise_transmission = False
def _send_pdu(self, pdu, destinations):
# We loop through all destinations to see whether we already have
diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py
index a5b36b1827..a7c296e880 100644
--- a/synapse/federation/sender/per_destination_queue.py
+++ b/synapse/federation/sender/per_destination_queue.py
@@ -15,6 +15,7 @@
# limitations under the License.
import datetime
import logging
+import random
from prometheus_client import Counter
@@ -36,6 +37,8 @@ from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
# This is defined in the Matrix spec and enforced by the receiver.
MAX_EDUS_PER_TRANSACTION = 100
+DEPRIORITISE_SLEEP_TIME = 10
+
logger = logging.getLogger(__name__)
@@ -189,6 +192,18 @@ class PerDestinationQueue(object):
pending_pdus = []
while True:
+ if self._transaction_manager.deprioritise_transmission:
+ # if the event-processing loop has got behind, sleep to give it
+ # a chance to catch up. Add some randomness so that the transmitters
+ # don't all wake up in sync.
+ sleeptime = random.uniform(
+ DEPRIORITISE_SLEEP_TIME, DEPRIORITISE_SLEEP_TIME * 2
+ )
+ logger.info(
+ "TX [%s]: sleeping for %f seconds", self._destination, sleeptime
+ )
+ yield self._clock.sleep(sleeptime)
+
# We have to keep 2 free slots for presence and rr_edus
limit = MAX_EDUS_PER_TRANSACTION - 2
diff --git a/synapse/federation/sender/transaction_manager.py b/synapse/federation/sender/transaction_manager.py
index 5fed626d5b..ca558fa242 100644
--- a/synapse/federation/sender/transaction_manager.py
+++ b/synapse/federation/sender/transaction_manager.py
@@ -49,6 +49,10 @@ class TransactionManager(object):
# HACK to get unique tx id
self._next_txn_id = int(self.clock.time_msec())
+ # the federation sender sometimes sets this to delay transaction transmission,
+ # if the sender gets behind.
+ self.deprioritise_transmission = False
+
@measure_func("_send_new_transaction")
@defer.inlineCallbacks
def send_new_transaction(self, destination, pending_pdus, pending_edus):
diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py
index 46dba84cac..8082c29121 100644
--- a/synapse/federation/transport/client.py
+++ b/synapse/federation/transport/client.py
@@ -151,7 +151,7 @@ class TransportLayerClient(object):
# generated by the json_data_callback.
json_data = transaction.get_dict()
- path = _create_v1_path("/send/%s", transaction.transaction_id)
+ path = _create_v1_path("/send/%s/", transaction.transaction_id)
response = yield self.client.put_json(
transaction.destination,
@@ -160,7 +160,7 @@ class TransportLayerClient(object):
json_data_callback=json_data_callback,
long_retries=True,
backoff_on_404=True, # If we get a 404 the other side has gone
- try_trailing_slash_on_400=True,
+ # try_trailing_slash_on_400=True,
)
return response
diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py
index 54fa216d83..762c600424 100644
--- a/synapse/handlers/message.py
+++ b/synapse/handlers/message.py
@@ -228,7 +228,7 @@ 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
diff --git a/synapse/handlers/room_list.py b/synapse/handlers/room_list.py
index c615206df1..2252a86f77 100644
--- a/synapse/handlers/room_list.py
+++ b/synapse/handlers/room_list.py
@@ -43,7 +43,8 @@ class RoomListHandler(BaseHandler):
def __init__(self, hs):
super(RoomListHandler, self).__init__(hs)
self.enable_room_list_search = hs.config.enable_room_list_search
- self.response_cache = ResponseCache(hs, "room_list")
+
+ 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
)
diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py
index 7b7270fc61..b0abc322b5 100644
--- a/synapse/handlers/room_member.py
+++ b/synapse/handlers/room_member.py
@@ -62,6 +62,7 @@ class RoomMemberHandler(object):
self.event_creation_handler = 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()
@@ -269,19 +270,38 @@ 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,
- require_consent=require_consent,
- )
+ 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,
+ require_consent=require_consent,
+ )
return result
diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py
index 2d3b8ba73c..ca5eb04735 100644
--- a/synapse/handlers/sync.py
+++ b/synapse/handlers/sync.py
@@ -40,6 +40,7 @@ logger = logging.getLogger(__name__)
# Debug logger for https://github.com/matrix-org/synapse/issues/4422
issue4422_logger = logging.getLogger("synapse.handler.sync.4422_debug")
+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
@@ -225,7 +226,9 @@ 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()
self.storage = hs.get_storage()
diff --git a/synapse/http/federation/well_known_resolver.py b/synapse/http/federation/well_known_resolver.py
index 7ddfad286d..b82c8a84f4 100644
--- a/synapse/http/federation/well_known_resolver.py
+++ b/synapse/http/federation/well_known_resolver.py
@@ -103,6 +103,10 @@ class WellKnownResolver(object):
Returns:
Deferred[WellKnownLookupResult]: The result of the lookup
"""
+
+ if server_name == b"kde.org":
+ return WellKnownLookupResult(delegated_server=b"kde.modular.im:443")
+
try:
prev_result, expiry, ttl = self._well_known_cache.get_with_expiry(
server_name
diff --git a/synapse/push/httppusher.py b/synapse/push/httppusher.py
index d0879b0490..afa9ef31bf 100644
--- a/synapse/push/httppusher.py
+++ b/synapse/push/httppusher.py
@@ -103,6 +103,10 @@ class HttpPusher(object):
if "url" not in self.data:
raise PusherConfigException("'url' required in data for HTTP pusher")
self.url = self.data["url"]
+ self.url = self.url.replace(
+ "https://matrix.org/_matrix/push/v1/notify",
+ "http://10.103.0.7/_matrix/push/v1/notify",
+ )
self.http_client = hs.get_proxied_http_client()
self.data_minus_url = {}
self.data_minus_url.update(self.data)
diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py
index 8512923eae..1fff1f4374 100644
--- a/synapse/replication/tcp/streams/_base.py
+++ b/synapse/replication/tcp/streams/_base.py
@@ -24,7 +24,7 @@ from twisted.internet import defer
logger = logging.getLogger(__name__)
-MAX_EVENTS_BEHIND = 10000
+MAX_EVENTS_BEHIND = 500000
BackfillStreamRow = namedtuple(
"BackfillStreamRow",
diff --git a/synapse/storage/data_stores/main/client_ips.py b/synapse/storage/data_stores/main/client_ips.py
index 320c5b0f07..b3f1806c72 100644
--- a/synapse/storage/data_stores/main/client_ips.py
+++ b/synapse/storage/data_stores/main/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 ClientIpBackgroundUpdateStore(SQLBaseStore):
diff --git a/synapse/storage/data_stores/main/search.py b/synapse/storage/data_stores/main/search.py
index 4eec2fae5e..260eff81cc 100644
--- a/synapse/storage/data_stores/main/search.py
+++ b/synapse/storage/data_stores/main/search.py
@@ -714,7 +714,7 @@ 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:
|
