Solution: Don't use syslog.
The FAQ actually has a section about this which suggests you to use
multilog in the daemontools-package instead.
I'm using syslog-ng, but I'm not under so much load that I could compare
that with other alternatives. Does anyone have experiences with syslog-ng?
Is it as bad as the original syslog?
Henrik.
At 07:15 PM 6/1/2000 +0200, you wrote:
>My mailserver runs an RedHat Linux based "homemade dist" OS.
>It's a pII 233 MHZ, and it's got 96 MB memory (+256mb swap).
>When qmail starts delivering, the syslogd daemon spikes to around 86% CPU,
>and it's nearly all taken by the system, not in userspace.
>
>My maillogs get around 25MB a week (maybe one can trim down that size a bit,
>by logging only errors in the "long" format ??).
>But when i tail -f /var/log/maillog i don't really think that it's logging
>that fast (even after a -ALRM signal to qmail-send with ~800 mails in the
>queue).
>So if i take like 25/7days/24 hours i get around 150KB logging per hour.
>That does not justify a 80% load of my system on syslogd's behalf...
>
>Does anyone recognize this, or even better have a solution?
>
>Any tip at all is super-welcome :)
>
>
>/Magnus Naeslund
