Hi Shkelzen,
> We have a strange behaviour with our wms+lb. Some jobs are just
> handled correctly but some other are staying in READY status for a
> long time. Then after between few minutes and some hours there are
> actually submitted.
Do those jobs have anything in common? E.g. destination CE(s) or user?
Can you provide the output of "glite-wms-job-logging-info -v 2" for
such a job?
> I can see that the input sandbox is correctly transfered.
>
> What could be wrong?
Do the WMS/LB file systems have enough free disk space and inodes?
Does your WMS/LB have enough memory? The WM deamon can take 2.5 GB
and might cause a lot of paging.
Is any WMS/LB daemon often being restarted? Check the start times of
processes with "ps auxw".
|