Hi folks. The problem was that syslogd on one of our 4.0G machines was
becoming hung occasionally, requiring a stop and start to get it logging
again.
Many thanks for invaluable and unflagging help from Jem Treadwell of
Compaq. The problem was tracked to a bug in libc, to do with reverse dns
lookup during a select call, as far as I understood.
Here is the combination of factors that caused the problem, and how we
fixed it: We had an empty syslog.auth file to disallow foreign logging to
our machine. There was a printer somewhere on our network, however, that
was configured to send its status messages to that machine. This printer
had an internal IP address, 192.168.2.2, but no reverse lookup. When it
tried to connect, syslogd was still trying to look up its name before
refusing access, and getting hung in the select() call while doing so
--sometimes. Syslogd got hung one time out of three in one of our tests.
We added reverse lookup for the printer's IP and enabled it in syslog.auth
to see what it was trying to say, and received some normal-looking printer
messages about paper being out, etc. Our syslogd has been fine since, no
hangs. It was an interesting set of events.
Ann Cantelow
Original question:
>
> Hi. We've been seeing a problem with syslogd logging in the past month or
> two on a Tru64 4.0G (unpatched) machine. At unpredictable times, syslogd
> suddenly stops all logging, though its daemon process remains up.
> Running stop and start of syslogd from /sbin/init.d starts the logging
> again. This happens several times a week, usually.
>
> We can't find anything in any of the log files that seems out of order.
> There is no coincidence that we can find of something running in
> cron. Searching patch release notes on compaq.com for 4.0G turns up
> nothing on syslogd.
>
> Has anyone seen something like this, or have ideas how I can track down
> what's happening?
>
> Thanks lots,
>
> Ann Cantelow
>
>
Received on Mon Nov 13 2000 - 16:33:44 NZDT