Re: mulitlog not getting terminated when Red Hat reboots?

On Mon, Jan 22, 2007 at 06:55:43PM -0600, Rob Mayoff wrote:
> >- What is 'outage', in the context of mulitlog?
> 
> It means that multilog died without a SIGTERM.  In your case, what's  
> probably happening is that multilog is exiting cleanly due to the  
> SIGTERM from shutdown, but then it's getting restarted by supervise  
> (which either hasn't gotten SIGTERM yet or has been restarted by  
> svscan which hasn't gotten SIGTERM yet).

Several people have brought up how svscan would affect me.  I should
have mentioned that I do have a SysV-style init script explicitly
that does do 'svc -d /service/foo/log', and there is the appropriate
'K23foo' symlink for my runlevel.

I can see that if I use my 'service foo stop' script from the command
line, my foo multilog process is indeed stopped.

  /service/foo/log: down 2 seconds, normally up

So - when I exit runlevel 3, my K23foo symlink to my 'foo' script
should be called.  svscan should not have restarted it.  'shutdown'
carpet-bombing the process tree with SIGTERM should not have caused
it to restart.

Ergo, I should have met the conditions for a clean termination of
multilog, and should have avoided this 'outage' condition.

Yet, it is happening.  One of my assumptions is being violated.

More testing, I guess.  I appreciate the feedback so far, thanks!

-- 
Brian Reichert				<[email protected]>
55 Crystal Ave. #286			Daytime number: (603) 434-6842
Derry NH 03038-1725 USA			BSD admin/developer at large