Wietse Venema:
> Patric Falinder:
> > Hi,
> >
> > Recently I've noticed that my mailqueue is filled up with doube-bounce
> > messages that won't go away.
> > I can't remember it being this way since the beginning and I just
> > noticed it a couple of days ago.
> >
> > Right now I have 193 messages in the queue and they're all from
> > double-bounce.
> > They are only in the queue for the time specified in
> > maximal_queue_lifetime that I have set to 3 days.
> > Before it was 5 days and then I had over 1000 messages in the queue..
> >
> > If I do a postcat on a message I get this:
> >
> > *** ENVELOPE RECORDS /var/spool/postfix/active/6322B10A46C ***
> > message_size: 256 236 1
> > 0 256
> > message_arrival_time: Sat Oct 9 13:59:08 2010
> > create_time: Sat Oct 9 13:59:08 2010
> > named_attribute: log_message_origin=local
> > named_attribute: trace_flags=256
>
> That is an address probe (triggered by the use of reject_unverified_sender
> or reject_unverified_recipient). Postfix does one delivery attempt
> and throws them away.
>
> Apparently your Postfix queued a lot of address probes a few days
> ago. If the messages are in the incoming or deferred queue, then
> your queue manager isn't waking up to grab new mail. This can be
> a symptom of broken timer support in virtualized environments.
Another observation:
Apparently the low-frequency $queue_run_delay (300s) timer still
works, so the queue manager picks up one message every 300s, but
your system's ZERO-DELAY timer support is broken, so the queue
manager does not pick up the next message immediately as it does
everywhere else.
Thus, you're limited to about 288 deliveries per day.
What OS platform is this? What about virtualization? What about
tick-less timer support? What Postfix version, any "improvements"
made by local users or helpful vendors? RedHat used to be pretty
good at breaking Postfix in surprising ways.
Wietse
