Xymon Mailing List Archive search

PID changes / Xymon restart

3 messages in this thread

list Thomas Seglard · Thu, 17 Feb 2011 13:24:53 +0100 ·
Hello,
 
I got a very strange behaviour on my platform with the Xymon client. The
cient process don't stop to change PID (and sometimes the process stops
running without any reasons), I could see this behaviour thanks to Monit
and here is an extract from the log :
 
CET Feb 14 14:28:29] info     : 'xymon-client' PID has not changed
[CET Feb 14 14:46:29] error    : 'xymon-client' process PID changed to
2687
[CET Feb 14 14:48:29] info     : 'xymon-client' PID has not changed
[CET Feb 14 15:08:29] error    : 'xymon-client' process PID changed to
21173
[CET Feb 14 15:10:30] info     : 'xymon-client' PID has not changed
[CET Feb 14 15:28:31] error    : 'xymon-client' process is not running
[CET Feb 14 15:28:31] info     : 'xymon-client' trying to restart
[CET Feb 14 15:28:31] info     : 'xymon-client' start:
/etc/init.d/xymon-client
[CET Feb 14 15:30:31] info     : 'xymon-client' process is running with
pid 7355
[CET Feb 14 15:50:32] error    : 'xymon-client' process PID changed to
24896
[CET Feb 14 15:52:32] info     : 'xymon-client' PID has not changed
[CET Feb 14 16:10:32] error    : 'xymon-client' process PID changed to
10575
[CET Feb 14 16:12:32] info     : 'xymon-client' PID has not changed
[CET Feb 14 16:30:33] error    : 'xymon-client' process PID changed to
28689
 
As a consequence each time PID changes, Puppet restarts Xymon client and
after few hours I see a lot of hobbitlaunch process confusing some of my
tests and filling up $XYMONHOME/tmp directory. I'm using Xymon for 4
years and never saw this before. My platform runs with Centos5.5 x86_64
and Xymon 4.2.3.
 
Why does Xymon change its PID like that ? Why does it suddenly stops
running (no trace of segfault in logs) ?
 
This problem is now very annoying thus any ideas or advices will be
appreciated J
Sincerly,
 
Thomas S
list Henrik Størner · Thu, 17 Feb 2011 13:44:08 +0000 (UTC) ·
quoted from Thomas Seglard
In <user-4b9d3f32f69f@xymon.invalid> "SEGLARD, Thomas" <user-da8ce279e795@xymon.invalid> writes:
I got a very strange behaviour on my platform with the Xymon client. The
cient process don't stop to change PID (and sometimes the process stops
running without any reasons)
First, I would like you to stop Monit/Puppet from restarting the client.
After that, kill all "hobbitlaunch" instances. Then start the client
with the "runclient.sh start" command.

That should start the client, and it should remain running until
explicitly stopped.

If it stops, there is a bug. What version of the client software
is this ? Older versions could behave oddly if the system clock
was stepped back in time - the 4.3.0 betas/RC's have this fixed.
As a consequence each time PID changes, Puppet restarts Xymon client and
"hobbitlaunch" will remain running with the same PID. The actual
client job - that is run by hobbitlaunch - will have a new PID
every 5 minutes.


Regards,
Henrik
list Thomas Seglard · Thu, 17 Feb 2011 17:01:32 +0100 ·
Hi Henrik,

Thanks for answer. I'm using the latest stable version : 4.2.3 for all my clients and server. I made what you asked me to do and as for now, everything is working. I will drop another mail in case I see clients suddenly stop.
Sincerly,

Thomas S
quoted from Henrik Størner

-----Message d'origine-----
De : Henrik "Størner [mailto:user-ce4a2c883f75@xymon.invalid] 
Envoyé : jeudi 17 février 2011 14:44
À : xymon at xymon.com
Objet : Re: [xymon] PID changes / Xymon restart

In <user-4b9d3f32f69f@xymon.invalid> "SEGLARD, Thomas" <user-da8ce279e795@xymon.invalid> writes:
I got a very strange behaviour on my platform with the Xymon client. The
cient process don't stop to change PID (and sometimes the process stops
running without any reasons)
First, I would like you to stop Monit/Puppet from restarting the client.
After that, kill all "hobbitlaunch" instances. Then start the client
with the "runclient.sh start" command.

That should start the client, and it should remain running until
explicitly stopped.

If it stops, there is a bug. What version of the client software
is this ? Older versions could behave oddly if the system clock
was stepped back in time - the 4.3.0 betas/RC's have this fixed.
As a consequence each time PID changes, Puppet restarts Xymon client and
"hobbitlaunch" will remain running with the same PID. The actual
client job - that is run by hobbitlaunch - will have a new PID
every 5 minutes.


Regards,
Henrik