Hi Maarten
Thanks a lot for helping me out of this. I can not confirm earlier as
batch system was in bad shape. In the continuation I would like to draw
your attention to earlier problem with pbs.pm
(https://gus.fzk.de/ws/ticket_info.php?ticket=31974 ) in which CE send
the GLOBUS_LOCATION to batch system. I solved this problem by creating a
sym link there but this is just hack.
Regards
Kashif
-----Original Message-----
From: LHC Computer Grid - Rollout [mailto:[log in to unmask]] On
Behalf Of Maarten Litmaath
Sent: 06 October 2009 22:22
To: [log in to unmask]
Subject: Re: [LCG-ROLLOUT] Job stay in running state
Hi Kashif,
> [...] no
> /opt/globus/libexec/grid_monitor_lite.sh process starts at CE when I
> submit job and /opt/globus/var/log/globus-gma.log is full of these
> warnings
>
> Fri Oct 2 16:05:04 2009:18593:WARN: Poll failed for job
> https://ngsce-test.oerc .ox.ac.uk:64007/15177/1254481706/ Fri Oct 2
> 16:05:04 2009:3873:WARN: Poll process terminated with error for job h
> ttps://ngsce-test.oerc.ox.ac.uk:64007/15177/1254481706/
>
> I could not found that why /opt/globus/libexec/grid_monitor_lite is
> not starting.
Understood now: your CE does not have the /usr/bin/file command!
In /opt/globus/lib/perl/Globus/GRAM/JobManager/fork.pm there is this:
----------------------------------------------------------------------
my $file_out = `/usr/bin/file $exec`;
if ( $file_out =~ /script/ || $file_out =~ /text/ ||
$file_out =~ m|/usr/bin/env| ) {
----------------------------------------------------------------------
The globus-gma/grid_monitor_lite functionality depends on the command,
but vdt_globus_jobmanager_common-VDT1.6.1x86_rhas_4_LCG-3 does not
require its presence; I will open a bug for that. In the meantime:
yum install file
|