Hi,
I know people on this list have suffered through this, so maybe
somebody can spell this out for me from experience. (T2K ;-) ?
I have a small VO (TM) who submits jobs to the grid using glite-wms-job-submit.
The jdl file looks like this (for the record I'd like to state that I
did not choose the name of the software myself):
Type = "Job";
JobType = "Parametric";
Executable = "run_mastercode_grid.sh";
StdOutput = "run_mastercode.log";
StdError = "run_mastercode.err.log";
InputSandbox = {"run_mastercode_grid.sh" , "config_grid.tar"};
OutputSandbox = {"run_mastercode.log" , "run_mastercode.err.log",
"mastercode.log", "make_masterfitter.log", "make_micromeags_only.log"
};
ShallowRetryCount = 5;
Requirements = (RegExp("ceprod05.grid.hep.ph.ic.ac.uk",other.GlueCEUniqueId) ||
RegExp("ceprod06.grid.hep.ph.ic.ac.uk",other.GlueCEUniqueId)) &&
(other.GlueCEPolicyMaxCPUTime > 1440);
Parameters = 100;
ParameterStart = 1;
ParameterStep = 1;
It's exactly the same job 99 times over, as it's basically Monte-Carlo
production.
Currently they mostly get away with a 24 h proxy made like this:
voms-proxy-init -valid 24:00 --voms vo.londongrid.ac.uk
Unfortunately some of the jobs need more time, either because they
queue for a while or because they've hit a particularly exciting bit
of phase space.
I can happily get a generic proxy with myproxy
lx05:grid_mastercode_nov11 :~] myproxy-info
username: dbauer
owner: /C=UK/O=eScience/OU=Imperial/L=Physics/CN=daniela bauer
retrieval policy: *
timeleft: 167:59:40 (7.0 days)
How do I get my jobs to renew their voms proxies (no voms proxy, no
access to our SE) in time ?
Thanks,
Daniela
--
-----------------------------------------------------------
[log in to unmask]
HEP Group/Physics Dep
Imperial College
Tel: +44-(0)20-75947810
http://www.hep.ph.ic.ac.uk/~dbauer/
|