Fix race condition in room stats. (#6029)

Broke in #5971

Basically the bug is that if get_current_state_deltas returns no new updates and we then take the max pos, its possible that we miss an update that happens in between the two calls. (e.g. get_current_state_deltas looks up to stream pos 5, then an event persists and so getting the max stream pos returns 6, meaning that next time we check for things with a stream pos bigger than 6)
pull/6056/head
Erik Johnston 2019-09-17 12:41:23 +01:00 committed by Richard van der Hoff
parent 1e19ce00bf
commit 70c52821ce
2 changed files with 11 additions and 4 deletions

1
changelog.d/6029.bugfix Normal file
View File

@ -0,0 +1 @@
Fix room and user stats tracking.

View File

@ -84,6 +84,13 @@ class StatsHandler(StateDeltasHandler):
# Loop round handling deltas until we're up to date # Loop round handling deltas until we're up to date
while True: while True:
# Be sure to read the max stream_ordering *before* checking if there are any outstanding
# deltas, since there is otherwise a chance that we could miss updates which arrive
# after we check the deltas.
room_max_stream_ordering = yield self.store.get_room_max_stream_ordering()
if self.pos == room_max_stream_ordering:
break
deltas = yield self.store.get_current_state_deltas(self.pos) deltas = yield self.store.get_current_state_deltas(self.pos)
if deltas: if deltas:
@ -94,7 +101,7 @@ class StatsHandler(StateDeltasHandler):
else: else:
room_deltas = {} room_deltas = {}
user_deltas = {} user_deltas = {}
max_pos = yield self.store.get_room_max_stream_ordering() max_pos = room_max_stream_ordering
# Then count deltas for total_events and total_event_bytes. # Then count deltas for total_events and total_event_bytes.
room_count, user_count = yield self.store.get_changes_room_total_events_and_bytes( room_count, user_count = yield self.store.get_changes_room_total_events_and_bytes(
@ -117,10 +124,9 @@ class StatsHandler(StateDeltasHandler):
stream_id=max_pos, stream_id=max_pos,
) )
event_processing_positions.labels("stats").set(max_pos) logger.debug("Handled room stats to %s -> %s", self.pos, max_pos)
if self.pos == max_pos: event_processing_positions.labels("stats").set(max_pos)
break
self.pos = max_pos self.pos = max_pos