"Todd A. Jacobs" <[EMAIL PROTECTED]> writes:
> I have a dedicated dual-processor server with 384 MB of RAM and a single
> SCSI drive. I'm running qmail and pop3d supervised according the LWQ,
> which sets softlimit to 2000000. Available inodes in the 2GB /var/qmail
> partition is 131,616, with split set to 23. Max file descriptors is 4,096.
> Local mail is being delivered to $HOME/Maildir/ on a seperate partition on
> the same drive, mounted sync (as is /var/qmail).
>
> As described in LWQ, I have remote concurrency set to 20 and local
> concurrency set to 10. Is this too low, given the specs? Or, considering
> the performance hit of running qmail-pop3d against a sync-mounted single
> drive, should I leave this alone?
"Too low" for the concurrencies would be relative to the amount of
mail you need to move; but your hardware should certainly handle more.
I supported concurrencies of 50 and 10 with no strain when my system
was a Cyrix 166+ with 96 meg of ram and IDE disks.
> If I *do* bump up the concurrency, what rule-of-thumb should I apply to
> softlimit? I don't really have a good feel for what the concurrency does
> to memory requirements. Do I even need to adjust it at all?
The softlimit is per process, so you don't need to think about it as
you change concurrencies. Well, you need to look at the average
actual use and maybe add more memory if you get to the point where
you're running out due to high concurrency.
> Basically, I have RAM and CPU cycles out the wazoo, but am a little
> constrained by drive speed and resources, and want to shuffle things in
> and out of the queue as quickly as possible so that there's room for the
> things that linger due to disk quota problems or whatever.
>
> All this assumes that Something Bad (tm) happens when the queue is filled
> (out of descriptors, inodes, or blocks). Maybe it doesn't--enlightenment
> is always welcome.
I think if you actually fill the queue that Something Bad does indeed
happen, though I've never gone there to investigate. Not *that* bad;
I imagine that incoming mail would not be accepted, which is better
than being accepted and lost. Of course if the condition persists
long enough that mail will be bounced by the server it's sitting on.
--
David Dyer-Bennet / Welcome to the future! / [EMAIL PROTECTED]
SF: http://www.dd-b.net/dd-b/ Minicon: http://www.mnstf.org/minicon/
Photos: http://dd-b.lighthunters.net/
