[GE users] sgeexecd doesn't know find libraries at boot time

Beadles, Jeff jeff_beadles at mentor.com
Mon Sep 10 21:01:09 BST 2007

    [ The following text is in the "iso-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

By default (and imo it's a horrible default), sge_execd inherits any environment variables set when it is run.
I would bet that you have an environment variable set, say LD_LIBRARY_PATH that contains the path to libimf.so.  When the system boots, it doesn't have this variable set, and fails your jobs.  When you restart it, it picks it up from your environment, and passes it thru to your jobs.
The fix, is to have your job script set LD_LIBRARY_PATH, or if you only have one platform type in your grid, you could use;
$ qsub -v LD_LIBRARY_PATH ...


From: Maxime Kinet [mailto:mkinet at ulb.ac.be]
Sent: Mon 9/10/2007 6:43 AM
To: users at gridengine.sunsource.net
Subject: [GE users] sgeexecd doesn't know find libraries at boot time

At boot time the nodes doesn't seem to know the path to the shared 
library. When launching a job I get the following error :

$path_to_openmpi/bin/mpirun: error while loading shared libraries: 
libimf.so: cannot open shared object file:
No such file or directory

The problem disappear when I restart sgeexecd on each node. Is there 
a way to avoid this restart?

thanks for helping.
Maxime Kinet
Université Libre de Bruxelles
Physique Statistique et Plasmas, CP 231
Campus Plaine - Boulevard du Triomphe,
1050 Bruxelles.

Tel.   : +32-2-650.59.08
e-mail : mkinet at ulb.ac.be

To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

More information about the gridengine-users mailing list