duplicate SYSLOG messages on grcond exit

[Denis Laplante]

   When grcond exits, a duplicate copy of old unix (kernel) messages
   appears in /usr/adm/SYSLOG.  The duplicates seem to include a complete
   retrospective of messages from the kernel since the last reboot.
   For example, I have copied here 2 sets of lines from /usr/adm/SYSLOG.

   NOTE: I am NOT looking for an explanation of the individual messages.
   I want to find out how to stop bogus duplicate messages from
   appearing in SYSLOG.

	grep -n "unix:" /usr/adm/SYSLOG
 61:Oct 26 18:38:57 beluga unix: core file access error
 84:Oct 29 11:46:06 beluga unix: NFS server ogopogo not responding still trying
 85:Oct 29 11:54:34 beluga unix: NFS server ogopogo ok
 86:Oct 29 13:29:26 beluga unix: NFS write error 13 on host ogopogo
 87:Oct 29 13:57:46 beluga unix: NFS write error 13 on host ogopogo
 92:Oct 29 17:18:48 beluga unix: core file access error
209:Oct 31 16:32:29 beluga unix: NFS write error 13 on host ogopogo
253:Nov  1 16:35:01 beluga unix: NFS server orca not responding still trying
256:Nov  1 17:03:30 beluga unix: NFS server orca ok
321:Nov  2 18:13:14 beluga unix: NFS server orca not responding still trying
341:Nov  5 11:37:11 beluga unix: NFS server orca ok
347:Nov  5 11:37:46 beluga unix: NFS server ogopogo not responding still trying
350:Nov  5 11:40:30 beluga unix: gfx: no keyboard
389:Nov  5 14:58:52 beluga unix: NFS server ogopogo ok


	Lines dating from a recent logout from the graphics console
Nov  7 17:05:35 beluga grcond[12193]: In limbo
Nov  7 17:05:36 beluga grcond[12193]: Alive
Nov  7 17:05:37 beluga grcond[12193]: CIO: gopogo ok
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS write error 13 on host ogopogo
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS write error 13 on host ogopogo
Nov  7 17:05:37 beluga grcond[12193]: CIO: core file access error
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS write error 13 on host ogopogo
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server orca not responding still trying
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server orca ok
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server orca not responding still trying
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server orca ok
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server ogopogo not responding still trying
Nov  7 17:05:37 beluga grcond[12193]: CIO: gfx: no keyboard
Nov  7 17:05:37 beluga grcond[12193]: CIO: NFS server ogopogo ok

   Note for example that the message 'core file access error' appears only
   once under the signature 'beluga unix:' (when it really happened), and 27
   times under 'beluga grcond[*]: CIO: '.  At every logout a new retrospective
   of unix messages get appended.  I have seen this on 3 different SGI
   machines (most print a full copy of the boot-time copyright notice), and
   don't believe this is correct behaviour.  Here is a procedure to 
   illustrate the problem on your machine: 

   (1) 'grep unix: SYSLOG' to show real error messages. Then pick key words
       from those error messages, for example:
   (2) "egrep -n 'IRIX|sc0|dks0|NFS|keyboard' SYSLOG"
       Where keyword 'IRIX' causes one line of the copyright notice to
       be printed, and the others correspond to some real errors picked
       up by 'grep "unix" SYSLOG' .

  Killing and restarting 'syslogd' does not affect this problem.  To examine
  memory of syslogd, I did 'killall -ABRT syslogd; strings ./core'.  No old
  messages were found.  I suspect a bug in kernel code for /dev/klog :
  where else but the kernel would those old messages be stored between exits
  of grcond?  By the way, my /etc/syslog.conf contains
	*.debug		/usr/adm/SYSLOG

Denis Laplante <laplante at ocgy.ubc.ca>