How to run a Panda job - example 1

WARNING: This page is obsolete.

Useful documents:

• Using DQ2
• DA on Panda

This is an example on how to setup and run a job on a grid using Panda when you have a package (in this case, TrigNtDecision) which resides in a different tree (in this case, Trigger/TrigAnalysis/TrigNtDecision).

Replace ~/Tutorial below with the ATLAS_TEST_AREA as defined in your ~/cmthome/requirements file. These instructions are for running on lxplus with 13.0.30; it should be similar to run at other places / kits.

Setup for each login

You will need to do this each time. Note the 32-bit mode used - there are problems with dq2 and python in 64-bit modes at the moment.

source ~/cmthome/setup.sh -tag=13.0.30,32

source /afs/cern.ch/project/gd/LCG-share/current/etc/profile.d/grid_env.sh

source /afs/cern.ch/atlas/offline/external/GRID/ddm/endusers/setup.sh.CERN

voms-proxy-init -voms=atlas:/atlas/ca

Checkout the software and compile

• Do the following to setup Panda:

mkdir ~/Tutorial/13.0.30
cd ~/Tutorial/13.0.30
cmt co PhysicsAnalysis/DistributedAnalysis/PandaTools
cd PhysicsAnalysis/DistributedAnalysis/PandaTools/cmt
source setup.sh
gmake

• Do this to compile your code/package (in this case, TrigNtDecision)

This is not a good example as this package does not have a version and will one day give errors when compiling with 13.0.30. I plan to replace this recipe soon. Note that you can replace this with your favourite package and job script and do the example; as long as you can compile and run locally, it should work with Panda. (Asoka, 25 Oct 2007)

cd ~/Tutorial/13.0.30
cmt co Trigger/TrigAnalysis/TrigNtDecision
cd Trigger/TrigAnalysis/TrigNtDecision/cmt
source setup.sh
gmake

• Create the run directory

mkdir ~/Tutorial/13.0.30/run

Run the job

Run the job; in the example below, test_BNL_job.py is the job script file.

• Setups (you need to do this for each new terminal session)

cd ~/Tutorial/13.0.30/Trigger/TrigAnalysis/TrigNtDecision/cmt
source setup.sh

cd ~/Tutorial/13.0.30/PhysicsAnalysis/DistributedAnalysis/PandaTools/cmt
source setup.sh

voms-proxy-init -voms=atlas:/atlas/ca

• Go to the run directory:

cd ~/Tutorial/13.0.30/run

• Now the command to submit to Panda is pathena; note that there is an input dataset specified as inDS - this is what will be used regardless of what is written in test_BNL_job.py.

pathena --inDS=misal1_csc11.005017.J8_pythia_jetjet.digit.RDO.v12003105 --outDS=user.AsokaDeSilva.misal1_csc11.v6 --nEventsPerJob=3 --split=1 --site=AUTO test_BNL_job.py

Remember to use your own name for the output dataset !

You should use the --site=AUTO flag or else the jobs will try to find data only at BNL.

For help on options, do

pathena --help

Results and Miscellaneous

• You can see your job from the Panda Monitor page; just type in your PandaID under the "Quick Search/Job" box on the left and click on the links to get your log files, data etc.

• To kill your job, run the following command. Note that it is the Job_ID and not the PandaID which is the argument to the following command.

pathena_util(Job_ID)

• If you store your results at TRIUMF, note that user data is purged (in ~ 1 week) so you should copy them elsewhere (see dq2 tools and dq2_get in particular). Information on how to retrieve your data should also be on the Panda Monitor page.

• Unless you are grid-savvy, you probably don't know where to send the data and/or your jobs. The CE and SE names and locations are stored in the TiersOfAtlas python dictionary, here: http://atlas.web.cern.ch/Atlas/GROUPS/DATABASE/project/ddm/releases/TiersOfATLASCache.py

-- DougSchouten - 07 May 2008 -- AsokaDeSilva - 18 Oct 2007