Edgaras Luko?evi?ius:
> mail-backend-1 postfix/cleanup[24838]: warning: unix_trigger_event: read
> timeout for service public/qmgr
> mail-backend-2 postfix/cleanup[14219]: warning: unix_trigger_event: read
> timeout for service public/qmgr
Wietse:
> Your active queue is full. The system is falling behind delivering mail.
Edgaras Luko?evi?ius:
> Default limit is:
> qmgr_message_active_limit = 20000
>
> I don?t have that many messages in active queue at any time.
>
> And the only way to avoid this warning is to either throttle new messages of
> try to deliver them faster?
Thanks for the additional information. The warnings mean that the
queue manager is blocked (not responding to trigger requests, which
are implemented as connections over UNIX-domain sockets), or that
those connection requests are being dropped in the kernel.
Things than can block the queue manager are:
- Mis-configuration (no retry service in master.cf). You would see
this as logging by qmgr with "status=deferred" (instead of such
logging by the error/retry service).
- Slow table lookups in trivial-rewrite that cause the queue manager
to fall behind on everything as it attempts to resolve a batch of
recipients from a multi-recipient message.
- Frequent use of "postfix reload" or "postfix flush" which can
make any performance problem worse.
Wietse
