Re: [hobbit] hobbit.sh not stopping server correctly

[Buchan Milne <bgmilne (at) staff.telkomsa.net>]

On Tuesday, 18 August 2009 12:25:26 j.sansford (at) ntlworld.com wrote:
> Hi Buchan,
>
> Apologies, I'll explain it in more detail. We are running on Solaris 10 and
> trying to configure Xymon as an SMF service (which essentially uses
> stop/start scripts). The stop script is basically:
>
> kill -TERM `cat /export/home/hobbit/logs/hobbitlaunch.pid`
>
> Where hobbitlaunch.pid is the PID of the hobbitlaunch process. When running
> this command we are left with many stray processes, not just a vmstat. For
> example:
>
> /usr/local/bin/ssh -l root 10.6.6.6 sysstat -m -c1 290
> sh -c (/usr/local/bin/ssh -l root 10.6.6.5 sysstat -x -c1 290
> >/export/home/hob
>
> Doing a bit more digging it seems these are related to an ext/netapp
> monitor perl script.

I haven't (yet) monitored any netapp devices with devmon (most likely I will 
soon) ... but there is a template, which has tests named 'cluster', 
'if_load','netapptemp','quota','status' and 'volume'. You may want to consider 
trying it instead of this script. If the teplate is missing something which is 
available by SNMP, we can add support for it.

BTW, are you running this out of cron, or from hobbitlaunch?

> Taking this out leaves less processes (although I
> don't want to leave it out forever so I'll have to do a bit of digging
> around to find out how to kill these gracefully) however because the
> iostats do not finish the SMF reports a failure rather than a disable when
> I try to stop the monitor.

Well, then it seems SMF is a bit brain-dead ... however, you could run 
something equivalent to:
pkill -u xymon vmstat

as well, which on my test box results in no processes running as the xymon 
user (after stopping the service via the init script on a Linux box).

> This means the SMF can't restart the bbserver if
> it ever crashes.

I don't think I've ever seen hobbitlaunch crash ...

Regards,
Buchan