syslog on sun{3,386i,4}

[Steve Losen]

In article <4244 at kalliope.rice.edu> metaware!riscy!adam at ucscc.ucsc.edu (Adam Margulies) writes:
>syslog seems to like to take over a system from time to time and eat 80%
>of the cpu time. It writes nothing to /usr/adm/messages so I can't tell if
>it is doing it for good REASON or anything.  syslog just hates me.
>
>how can i free myself from it, other than turning it off all together?
>
>-- 
>adam margulies          metaware inc.         "compilers, from a higher source"
>
>UUCP: [...]!ucbvax!sco.com!metaware!adam               ATT: (408)429-META x3016

I noticed this too.  The problem can be fixed in the /etc/syslog.conf
file.  I commented out the following line and this stopped the problem.

mail.debug                     ifdef(`LOGHOST', /var/log/syslog, @loghost)

Note that LOGHOST was not defined and that in /etc/hosts there was a
loghost alias for the local host.

In case anyone cares, I discovered that sendmail was triggering syslogd to
loop.  Upon further investigation I found that the following command
sequence also caused syslogd to loop.

logger -p mail.err
logger -p mail.info

As a matter of fact, any two logger commands with any two different mail.*
priorities triggered the looping.  Non-mail priorities had no effect.
Repeating the same "logger -p mail.*" command had no effect.

By the way, this happened with release 4.0.1 on both a sun3/60 and a
sun4/260.

Steve Losen     scl at virginia.edu
University of Virginia Academic Computing Center