PandaTestJob
Introduction
This page describes how to test
Panda jobs.
Production managers or ADCoS shifters should if possible use the test interface available on
atlas002.uta.edu. This allows to use an uniform environment and avoids the need of new installations. Check
Setup atlas002 section for details on the installation on atlas002.
Setup
$ mkdir panda
$ cd panda
For SL(C)5
$ svn co svn+ssh://svn.cern.ch/reps/panda/panda-server/current/pandaserver panda-server
For SL(C)6
$ svn co svn+ssh://svn.cern.ch/reps/panda/panda-server/branches/sl6/pandaserver panda-server
$ cd panda-server
$ export PYTHONPATH=$PWD:$PYTHONPATH
$ cd test
Note that you need a grid proxy and Python 2.3 or higher.
Setup on atlas002
For shifters with access to atlas002. The software to submit Panda test jobs on
atlas002 is installed in
/atlas-grid/pandatest_current
You can log in the host, change to the directory, modify the PYTHONPATH and submit the test jobs.
If you change some of the test jobs please update the CVS version.
Send jobs
SYNOPSIS
python testXYZ.py [siteID]
where testEvgen.py for evgen, testG4Sim.py for simul, and testReco.py for reco.
For example,
$ python testReco.py
If you want to send jobs to a specific site,
$ python testReco.py BU_ATLAS_Tier2
If you send jobs to a specific site, pay attention that the Cloud is set correctly in the python file defining the test job.
E.g. if you send a testSimulReco14 job to BU_ATLAS_Tier2 to allow the correct delivery of input files you should have in testSimulReco14.py:
job.cloud = 'US'
You may fine in the same directory testXYZNN.py, where NN is 14,15,16,17. For example, testG4sim17.py is for ATLAS release 17.
If you want to test another release/cache, change the following two lines in testXYZNN.py:
job.AtlasRelease = 'Atlas-17.0.5'
job.homepackage = 'AtlasProduction/17.0.5.6'
For testG4sim, you can specify dataset name, input file name, cloud, and site name in the command line.
python testG4sim17.py [siteID] [cloud] [dataset] [filename]
Monitoring test jobs
Note that test jobs do not show up in the regular monitoring pages like production dashboard. You have to check specifically for test jobs, e.g.:
http://panda.cern.ch:25880/server/pandamon/query?job=*&type=test
Kill jobs
SYNOPSIS
python killJob.py pandaID1 [pandaID2]
Kill 123
$ python killJob.py 123
Kill 123-135
$ python killJob.py 123 135
Reassign jobs
SYNOPSIS
python reassignJobs.py PandaID1 [PandaID2]
Reassign 123
$ python reassignJobs.py 123
Reassign 123-135
$ python reassignJobs.py 123 135
Release waiting jobs
SYNOPSIS
python reassignWaiting.py
Major updates:
--
TadashiMaeno - 17 Jan 2006
Responsible: TadashiMaeno