syslogd not draining
Michael W. Lucas
mwlucas at blackhelicopters.org
Mon Mar 27 19:07:00 UTC 2006
On Mon, Mar 27, 2006 at 10:59:16PM +0400, Maxim Konovalov wrote:
> [...]
> > > ~100 datagrams per second, not a lot. Perhaps they are huge.
> >
> > Not that I've noticed. It's syslogd, DHCP, DNS, and flow-capture
> > from a variety of devices, all generally small packets.
>
> I know nothing about your netflow setup but it can generate a
> huge amount of udp traffic. Check 'pkts' and 'lost' fields in
> /var/log/flow-capture.
>
> > > > > How much cpu time does syslogd use?
> > > >
> > > > Not much. ps -ax | grep syslog gives:
> > > >
> > > > 4317 ?? Ss 0:01.60 /usr/sbin/syslogd -l /var/run/log -l
> > > > /var/named/var/run/log
> > >
> > > Try to remove a log socket for named and restart syslogd.
> >
> > Removed the named socket and restarted. We'll see what happens.
> [...]
>
> Can you check there is nothing from syslogd on the serial console?
I changed that this morning in response to an earlier suggestion.
To be sure, I just ran:
logger -p local0.err "test message"
and got a message in /var/log/console.log, nothing on the serial console.
BTW, running syslogd with -n seems to help the problem -- not cure it,
mind you, but reduces its severity.
==ml
--
Michael W. Lucas mwlucas at FreeBSD.org, mwlucas at BlackHelicopters.org
http://www.BlackHelicopters.org/~mwlucas/
"The cloak of anonymity protects me from the nuisance of caring." -Non Sequitur
More information about the freebsd-hackers
mailing list