retention purge job don't run

[4nd3r]

Steps to reproduce

• set retention: enabled: true in homeserver config
• restart synapse
• send event m.room.retention with content { "max_lifetime": 86400000 }
• select * from room_retention confirms that message retention is configured for room
• wait 24 hours and bit more

  synapse/synapse/config/server.py

  Line 368 in c2ba994

  "interval": self.parse_duration("1d"),

• messages older than 86400000 ms still in the room (even after clear cache and reload)

Version information

• Homeserver: private
• Version: 1.8.0
• Install method: virtual env & pip
• Platform: debian 9 with postgresql 9.6

[babolivier]

Thanks for the report, I will have a look later next week. FWIW, can you find lines which message start with [purge] in your Synapse logs?

[4nd3r]

FWIW, can you find lines which message start with [purge] in your Synapse logs?

no such messages in log (loglevel=INFO).

[babolivier]

Hmm, weird. Even if the purge job doesn't run, Synapse should prevent expired events from being sent to clients:

synapse/synapse/visibility.py

Lines 125 to 135 in ac87ddb

	# Don't try to apply the room's retention policy if the event is a state event, as
	# MSC1763 states that retention is only considered for non-state events.
	if apply_retention_policies and not event.is_state():
	retention_policy = retention_policies[event.room_id]
	max_lifetime = retention_policy.get("max_lifetime")
	if max_lifetime is not None:
	oldest_allowed_ts = storage.main.clock.time_msec() - max_lifetime
	if event.origin_server_ts < oldest_allowed_ts:
	return None

Therefore events shouldn't be visible from Riot after a "Clear cache and reload"...

I've tried to reproduce this in a room with the same settings as yours, with a few messages in there:

and then waited a bit more than a day and hit "Clear cache and reload" in Riot, and the messages had disappeared as expected:

So it looks to me like Synapse doesn't believe the feature has been enabled on your server, which leaves me puzzled, as it shouldn't happen if you have

retention:
    enable: true

in your homeserver's configuration :/

[babolivier]

At this point the only thing I can think of that might help is restart your homeserver to ensure it registered the configuration change...

[richvdh]

@babolivier: would it be worth adding some more logging to help us see what purge jobs are being set up, and when they are running?

[babolivier]

@babolivier: would it be worth adding some more logging to help us see what purge jobs are being set up, and when they are running?

Oh good idea, I'll look at that tomorrow!

[4nd3r]

upgraded to 1.9.0 and first line in logs was:

Setting up purge job with config: {'longest_max_lifetime': None, 'shortest_max_lifetime': None, 'interval': 86400000}

should I turn loglevel to DEBUG or is INFO enough?