Re: Solaris 2.6 syslod dies on SIGHUP

[john_smith () rd qms com]

          On a partially related note (and word of warning) daemons in 
     Solaris 7 occassionally die for no apparent reason.  No warnings or 
     errors in the logs or on the console they just die.
     
          We've seen inetd, syslogd and named (public BIND, not the bundled 
     named) die on different Solaris 7 machines but haven't seen any 
     symptoms to try and track down.
     
     Thanks,
     
     John Smith


______________________________ Reply Separator _________________________________
Subject: Solaris 2.6 syslod dies on SIGHUP 
Author:  <Kenneth_W_Fox () sbphrd com> at Internet-Mail
Date:    7/23/99 10:15 AM


Anyone got a fix for the following:
     
     syslogd stops logging to a messages file when it is sent a SIGHUP,
such as when the logs are rotated. The process itself remains running, but 
nothing is written to the file. The only wey to get it going again is to 
stop it (kill pid) and restart it. The problem is that the on restart, 
syslogd reports all the Kernel debug information & it looks like the system 
has been rebooted. Apparent reboot == Manager coronary.  The system 
involved is a Gauntlet firewall V5.0.
     
     I checked out Sunsolve & tried thier patch (104639-04) and this didn't
fix the problem. The current work around was to alter the  log rotation 
script to
     
          /etc/rc3.d/S74syslog stop
          rotate logs
          /etc/rc3.d/S74syslog start
     
     Obviously this is not elegant or desireable. Our next thought is to
change the log rotation script so that it moves messages.0 through 
messages.nn to messages.(0+1) through nessages(nn+1) and to COPY  the 
messages file to messages.0 then copy /dev/null to the messages file.
     
     Has anyone else seen this or resolved in a more effective manner??
     
Thanks, Ken