Skip to content
This repository has been archived by the owner on Apr 26, 2024. It is now read-only.

Fix race calling /members?at= #14817

Merged
merged 2 commits into from
Jan 12, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions changelog.d/14817.bugfix
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Fix race where calling `/members` or `/state` with an `at` parameter could fail for newly created rooms, when using multiple workers.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it just newly-created rooms that are vulnerable to this race? E.g. are there other consumers of get_last_event_in_room_before_stream_ordering?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm, it can also be a bit weird if it races with any state change, as it may return the state before the state change. And I guess joins would also do something weird here.

Will update.

65 changes: 59 additions & 6 deletions synapse/storage/databases/main/stream.py
Original file line number Diff line number Diff line change
Expand Up @@ -801,13 +801,66 @@ async def get_last_event_in_room_before_stream_ordering(
before this stream ordering.
"""

last_row = await self.get_room_event_before_stream_ordering(
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is get_room_event_before_stream_ordering just wrong? Should we be auditing its other call sites?

room_id=room_id,
stream_ordering=end_token.stream,
def get_last_event_in_room_before_stream_ordering_txn(
txn: LoggingTransaction,
) -> Optional[str]:
# We need to handle the fact that the stream tokens can be vector
# clocks. We do this by getting all rows between the minimum and
# maximum stream ordering in the token, plus one row less than the
# minimum stream ordering. We then filter the results against the
# token and return the first row that matches.

sql = """
SELECT * FROM (
SELECT instance_name, stream_ordering, topological_ordering, event_id
FROM events
LEFT JOIN rejections USING (event_id)
WHERE room_id = ?
AND ? < stream_ordering AND stream_ordering <= ?
AND NOT outlier
AND rejections.event_id IS NULL
ORDER BY stream_ordering DESC
) AS a
UNION
SELECT * FROM (
SELECT instance_name, stream_ordering, topological_ordering, event_id
FROM events
LEFT JOIN rejections USING (event_id)
WHERE room_id = ?
AND stream_ordering <= ?
AND NOT outlier
AND rejections.event_id IS NULL
ORDER BY stream_ordering DESC
LIMIT 1
) AS b
DMRobertson marked this conversation as resolved.
Show resolved Hide resolved
"""
txn.execute(
sql,
(
room_id,
end_token.stream,
end_token.get_max_stream_pos(),
room_id,
end_token.stream,
),
)

for instance_name, stream_ordering, topological_ordering, event_id in txn:
if _filter_results(
lower_token=None,
upper_token=end_token,
instance_name=instance_name,
topological_ordering=topological_ordering,
stream_ordering=stream_ordering,
):
return event_id

return None

return await self.db_pool.runInteraction(
"get_last_event_in_room_before_stream_ordering",
get_last_event_in_room_before_stream_ordering_txn,
)
if last_row:
return last_row[2]
return None

async def get_current_room_stream_token_for_room_id(
self, room_id: str
Expand Down