Re: [exim] exim performance: how to send 10^6 letters daily?

Etusivu
Poista viesti
Vastaa
Lähettäjä: Ian Eiloart
Päiväys:  
Vastaanottaja: Peter Volkov
Kopio: exim-users
Aihe: Re: [exim] exim performance: how to send 10^6 letters daily?
I'm bringing this back to the list, for everyone's benefit.

--On 28 April 2009 21:25:55 +0400 Peter Volkov <volkov.peter@???>
wrote:

> Thank you Ian and Marc.
>
> В Втр, 28/04/2009 в 12:24 +0100, Ian Eiloart пишет:
>> Finally, how do you run your queue runners?
>
> Ian, I'm just starting main exim process as
>
> /usr/sbin/exim4 -bd -qq2m
>
> Previously I have -q2m but now I followed your suggestion and added
> another q to make two pass run.
>
>> Don't have too many running at one time, and try using a two pass run.
>
> This suggestion a bit contradicts to what I've read on this mailing
> list. I thought that the more queue runners I have running the faster
> mail will leave queue. Also I've read on this mailing list suggestion to
> upper queue_run_max (and I've put it equal to 100). But as I see from
> the logs some queue runners (qr) step on other qr messages:
>
> 2009-04-28 20:29:56 1Lyotv-00043M-2D Spool file is locked (another
> process is handling this message)
>
> And again as I found in this mailing list, this is merely a warning and
> I should take it seriously only if this message is repeated for the same
> message many times. So how many queue runners should be started?
> Currently queue processing is really slow and I'm not sure how to make
> it faster. Any way currently 144 is running on the server and it's
> slow... Probably I should decrease number of queue runners?


The most important thing is to keep the queue smaller.

> В Втр, 28/04/2009 в 07:19 -0700, Marc Perkel пишет:
>> What I've had luck with is putting the queues in a ram disk. Then
>> setting it up with a fallback host so that if the message isn't
>> delivered on the first try it transfers to another exim server that
>> keeps the queues on a hard drive.
>
> Interesting. We'll increase the RAM to 8G too and mount /var/spool/exim
> to tmpfs. But have you tried such setup without fallback host? I
> envision that it's possible that disk became overfilled. What happens
> next? Will exim stop accepting messages?




--
Ian Eiloart
IT Services, University of Sussex
01273-873148 x3148
For new support requests, see http://www.sussex.ac.uk/its/help/