Re: Behaviour change request.

[Jan-Henrik Haukeland]

Eetu Rantanen <address@hidden> writes:

> > I'm not sure why you was able to actually kill this Java process.
> > Maybe it wasn't started from within monit in the first place and it
> > was a light-weight process (thread) in the JVM that become a zombie
> > (not unusual) and when you killed (without -9) the JVM gracefully
> > waited for all of it's threads and thereby removed the zombie from the
> > system?
> 
> I forgot to mention, that the process was running under Solaris 8. The PID
> that was claimed to be a zombie was a single javaprocess (not a thread
> that had been forked by a parent-process).
> 
> Next time, we'll have to try and gdb/strace it, but I wasn't too bright at
> 2.30AM ;-)
> 
> What also came to my mind (And what Jani suggested earlier) is that if
> this is the wanted behaviour, why not add a statement such as alertmax or
> similar that would limit the amount of alerts monit sends.


As I said in another mail, it's clearly wrong for monit to send 200
alerts when a process has become a zombie. Enclosed is a patch that
will raise one alert and not check the process anymore.

Installation: 
Copy to the monit src directory and run, "patch < zombie.diff"

zombie.diff
Description: zombie.diff

-- 
Jan-Henrik Haukeland