[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

19991014: LDM log file problem (cont.)

>From: "Jason J. Levit" <address@hidden>
>Organization: Center for Analysis and Prediction of Storms, University of Okla
homa
>Keywords: 199910111619.KAA28000 LDM syslog

Jason,

re: exactly what happened
>Sorry for being cryptic! :-)  The above is correct, that is, I deleted
>all of the old log files, simply did a "touch ldmd.log" - to create a 
>blank file - and then typed "ldmadmin newlog".  What I assume should
>have happened is that ldmd.log should have been moved to ldmd.log.1, and a
>new ldmd.log file created.  This did not happen, nothing changed in the 
>"logs" directory.

OK.  The only thing is that an 'ldmadmin newlog' will not rotate the files
if ldmd.log is zero length.  Creating it with a touch makes it, but it
will be zero length.

re: group for user running LDM could have changed
>That's a good point, the user group id for "ldm" could have changed,
>and I checked it out...all looks good.  The group id is the same.

OK.

re: did syslogd have to be restarted?
>From the looks of things, syslogd was running just fine.  I simply had
>the sysadmin restart it.

OK.

re: quick look around
>Sure, if you want to give it a shot, that's great.  I'm dumbfounded at
>this point:

I logged on and found that there are _two_ copies of syslogd running:

ldm@dot% ps -eaf | grep syslogd
    root  1050     1  0   Sep 07 ?       327:28 /usr/sbin/syslogd
     ldm  8614  8328  0 17:46:00 pts/1    0:00 grep syslogd
    root  5869     1  0 15:22:55 ?        0:00 /usr/sbin/syslogd

This is not good.  Your system administrator must have not killed
the original one.  The fact that /var/adm/messages is not being written
to is telling us that the original syslogd is not functioning properly.
So, the procedure is to have 'root':

o kill both copies of syslogd that are running
o start a new copy of syslogd and verify that logs to /var/adm/messages
  and to ~ldm/logs/ldmd.log

I verified that a 'ldmadmin newlog' will work by writing some junk into
the zero length copy of ~ldm/logs/ldmd.log and then running 'ldmadmin newlog':

ldm@dot% ls -l
total 0
-rw-r--r--   1 ldm      craft          0 Oct 14 17:42 ldmd.log
ldm@dot% echo lasdfladf > ldmd.log
ldm@dot% ls -l
total 2
-rw-r--r--   1 ldm      craft         10 Oct 14 17:55 ldmd.log
ldm@dot% ldmadmin newlog
ldm@dot% ls -l
total 2
-rw-r--r--   1 ldm      craft          0 Oct 14 17:55 ldmd.log
-rw-r--r--   1 ldm      craft         10 Oct 14 17:55 ldmd.log.1

>Thanks for checking all of this out.  This one is a weird mystery!

So, it looks like the problem really is being caused by Sun's bug with
syslogd. Geez...

Tom

>From address@hidden  Fri Oct 15 08:43:49 1999

re: two copies of syslogd running
>Yikes, indeed you are correct!  I'll let the system administrator know,
>but it might not happen today...I believe he's on his way back from
>Seattle.

>I'll let you know if that fixes the problem; sounds pretty likely.

>Thanks again for all of your help and troubleshooting.  Weird problem!

Jason