Herisanu, good day.
Fri, Sep 05, 2008 at 01:14:34AM +0300, Herisanu Alexandru wrote:
> I struck a problem with the recent site reconfiguration. Is is a Torque or
> Maui related problem.
>
>
>
> [root@gw02 CONFIG]# tail -f /var/log/maui.log | grep ERROR
>
> 09/05 00:15:37 ERROR: cannot get server info: Premature end of message
>
> 09/05 00:15:58 ERROR: cannot get server info: Premature end of message
>
> 09/05 00:50:40 ERROR: cannot get node info: Premature end of message
And what you'll got if 'showq' is spawned during the periods when Gstat
gives you an error? Also try 'strace -p <Maui PID>' during that time.
If strace will show you that Maui performs a read() syscall, then you're
probably hitting the bug described in
http://www.clusterresources.com/pipermail/mauiusers/2008-June/003381.html
If you are hitting this problem, try to make Maui PBS query timeouts
to be larger as described in
http://www.clusterresources.com/pipermail/mauiusers/2008-June/003388.html
I had not yet settled the patch for this issue with Torque/Maui
developers, so it is not yet in the released versions.
--
Eygene Ryabinkin, Russian Research Centre "Kurchatov Institute"
|