[Mailman-Users] locks, max cpu, postfix, qrunner ugliness

Morgan Fletcher morgan at hahaha.org
Mon Feb 18 04:35:14 CET 2002


Charlie Watts <cewatts at frontier.net> writes:
>> > I think our list is hitting some sort of postfix limitation, but I'm not
>> > sure what it is.
>
> Why do you think it's postfix? That doesn't make sense, based on what
> you've said.

Yeah, it's python/qrunner that's wedged, postfix seems OK.

> Anyway, I'd ktrace/strace that python instance, see what it's doing. And
> look in your mail logs - is it actively dumping things into postfix, or
> has it not even gotten there, or has it started and paused?
>
> Is python just sitting there busy-waiting?

I 'ktrace -dp pid', waited five minutes, did 'ktrace -c', then 'kdump |
less'.

It's just an endless stream of:

  7272 python   CALL  break(0x9892000)
  7272 python   RET   break 0
  7272 python   CALL  break(0x9892000)
  7272 python   RET   break 0
  7272 python   CALL  break(0x96a6000)
  7272 python   RET   break 0
  7272 python   CALL  break(0x96a6000)
  7272 python   RET   break 0
  7272 python   CALL  break(0x9892000)
  7272 python   RET   break 0

I also should mention that I tested this mailman installation for about a
month with a small list (~30 addresses) with no problems, and for about a
week with a larger list (~400 addresses), also with no problems. Both those
lists contained valid addresses. We did test subscribing an address to the
second list, then removing the user that was subscribed at the user's host,
and mailman successfully unsubscribed it after it bounced for a while.

Now the 1800-member list is generating a ton of bounces, and I have a
feeling that's what's choking qrunner. But I may have been biased by
reading this:

http://sourceforge.net/tracker/?group_id=103&atid=100103&func=detail&aid=449550

Morgan




More information about the Mailman-Users mailing list