Running
SAM tests is restricted to people with the lcgadmin VOMs role. The DIRAC3 SGMEnv script should be used to generate the environment for the procedures outlined below.
Submitting to all sites
There is a pre-prepared DIRAC script for submitting
SAM jobs to all sites.
$ dirac-lhcb-sam-submit-all
If problematic sites are found, then they can be excluded from the daily
SAM test by adding the DIRAC site name to the following CS option:
/Operations/SAM/BannedSites
Tests are normally run by hand for CERN.
Targetting one site
If a job fails at a site, you may want to target it specifically to find out more about what is going on.
$ dirac-lhcb-sam-submit-ce
To get the CE of the site, use this tool to query the CS:
$ dirac-admin-site-info LCG.UKI-SCOTGRID-ECDF.uk
In case of shared area corruption, the targeted job should specify:
- CE of the site
- The existing lock file should be removed (--removeLock=True)
- Deletion of the shared area (--deleteSharedArea=True)
This will help to correct the problem. Only one such job should be sent at a time to prevent both jobs being able to write to the shared area simultaneously.
--
GreigCowan - 21 Sep 2008