[15093] in Athena Bugs
syslogd lossage on qs-3
daemon@ATHENA.MIT.EDU (John Hawkinson)
Sat Apr 19 01:33:24 1997
Date: Sat, 19 Apr 1997 01:33:15 -0400
To: bugs@MIT.EDU
Cc: quickcomments@MIT.EDU, jweiss@MIT.EDU
From: John Hawkinson <jhawk@MIT.EDU>
[[ I mentioned this to jweiss and he suggested I should report it,
as some people might have been deluded into thinking the problem
was fixed or had gone away of it's own accord. ]]
It appears that sometimes the Solaris syslogd fails to start up
properly at boot time, and never logs anything to any of it's logfiles.
I observed this on the quickstation where I sent mail to quickcomments
a few nights ago -- after sending my mail I wanted to login and check
the /usr/spool/mqueue/syslog log, and was surprised to find no entries.
I tried something like
logger -p mail.info foo
and found it didn't log.
I killed syslgod and restarted it with -d (debug) and it performed
normally (logger worked).
I killed it again and restarted it without debugging and it performed
normally (logger worked).
This is consistent with jweiss's previously reported experiences, viz.
that it works fine once restarted.
Perhaps this is some sort of boot-time lossage.
I don't believe this has ever been exhibitted with the syslogd implementation
that I wrote Solaris kernel support for, viz. the one in the Zephyr source
tree.
I'd suggest that people with private Suns that might exhibit this problem
should be encouraged to start syslogd at boot time with -d and direct
output to a file. Do this by modifying /etc/init.d/syslogd.
I'd consider doing this on the SIPB sparcs but they've never seen this
particular lossage (afaik), so I wonder if there's something special.
I know Jonathon has seen it multiple times.
This problem certainly needs to be investigated and barring any better
conclusion, submitted to Sun.
--jhawk
