[Mailman-Users] cpu usage
jonc at nc.rr.com
Sun Feb 15 02:34:07 CET 2004
There are a lot of items that can cause this.
- Using "top" look at the qrunners process ID's and see which ones are
eating the most cpu time.
- Now use "ps aux" on a *wide* terminal so you can see which qrunner is
associated with the high cpu process ID
Each of the qrunners is launched with its specific job:
qrunner --runner=CommandRunner:0:1 -s
qrunner --runner=IncomingRunner:0:1 -s
qrunner --runner=NewsRunner:0:1 -s
qrunner --runner=OutgoingRunner:0:1 -s
qrunner --runner=VirginRunner:0:1 -s
What is your Qrunner_Sleep_Time set to in ~mailman/Mailman/mm_cfg.py (or
Defaults.py)? The default of 1 normally works well.
QRUNNER_SLEEP_TIME = seconds(1)
Are you getting any errors in any of your log files when Mailman is
crunching on the mails? Once the mails are crunched and processed, do
the qrunner processes stay high, or drop down again?
If I can get enough data points on this problem I'll try to write up an
FAQ to cover it and the various places to explore.
On Sat, 2004-02-14 at 13:22, Adam Kessel wrote:
> > I have scoured the archives and google for answer to my problem but
> > none of the fixes I have found seem to work. I have mailman 2-1.1-91
> > installed w/ postfix 1.1.12-12 on a Suse 8.2 box and qrunner/python is
> > eating the cpu. Most of the posts I read describe this as a postfix
> > problem and to change my local delivery code to 550 which I did but it
> > hasn't helped. The system runs fine for a while but after a few hours
> > qrunner/python cpu usage goes through the roof. In most cases this is
> > with less than five emails in the queue.
> I've got a similar situation here. I'm running Debian sid: mailman
> 2.1.4-1 and postfix 1.1.11-0.woody3. qrunner CPU usage seems to go
> between 50-90%, and I'm not running any large lists (either in
> subscribers or frequency of postings). The machine should have
> sufficient resources: 200M RAM and a PII 450MHz CPU.
> Any ideas how to make mailman load more manageable?
> I also noticed this Debian bug report, which suggests that there's
> nothing that can be done:
More information about the Mailman-Users