Using PanDA? Athena

1 Atlas TWiki Links

The PanDA? Production and Distributed Analysis System: https://twiki.cern.ch/twiki/bin/view/Atlas/Panda

Client tools for Panda analysis jobs: https://twiki.cern.ch/twiki/bin/view/Atlas/PandaTools

How to submit Athena jobs to Panda: https://twiki.cern.ch/twiki/bin/view/Atlas/PandaAthena

2 Pathena Full Chain

Used Athena Version: 14.5.0.2

Jobs in pathena are run just as any job in Athena on a JobOptions file:

> athena.py <JobOptions>

using the pathena command on a Job Options file. OutputDataset and InputDataset is all that you have to provide in addition:

> pathena <JobOptions>
       [--inDS <InputDataset>]
       --outDS <OutputDataset>

Pathena copies the contents of the entire directory it is ran from to the remote site.

2.1 Generation

More information can be found here: https://twiki.cern.ch/twiki/bin/view/Atlas/WorkBookGeneration
General pathena example here: https://twiki.cern.ch/twiki/bin/view/Atlas/PandaAthena#example_1_evgen

2.1.1 How to get Job Options files:

You need to write or get the JobOptions you run pathena on. You can choose from a variety of files available on http://reserve02.usatlas.bnl.gov/lxr/source/atlas/Generators/EvgenJobOptions/share/:

• MC8.105145.PythiaZmumu.py for Z->mu,mu decay,
• MC8.105144.PythiaZee.py for Z->e,e decay,
• and many others

Use the following command to get Job Options you want, we are going to use the Z->e+,e- decay:

> get_files -jo MC8.105144.PythiaZee.py

2.1.2 How to change minimum number of events:

The default value of the MC8.105144.PythiaZee.py is 5000 events, therefore if you choose your below 5000, you will get into problems and generation will crash.

If you want to change this, you need to edit the JobOptions? .py file (e.g. MC8.105144.PythiaZee.py) and add this line to the end:

evgenConfig.minevents = 100 # default is 5000

For pathena we can of course leave the default 5000, since pathena standard pathena jobs get killed after 90 minutes rather than 45 minutes on the LXPLUS. (pathena --long allows you to submit to a long queue).

2.1.3 Running Pathena on JobOptions?

is basically the same as running athena on JobOptions:

> pathena pomwig.py --outDS user09.MartinZeman.pomwig.evgen.pool.v14000502 --noBuild

2.1.4 Running Generation using CSC Job Transformation on Pathena

Locally you can run Evgen using Job Transformation by issuing:

> csc_evgen08_trf.py <RunNo.> <FirstEvent> <MaxEvents> <RandomSeed> ./<JobOptions.py> <Output.evgen.pool.root>

In our case, we have MC8.105144.PythiaZee.py so the transformation will be locally run like this:

> csc_evgen08_trf.py 105144 1 110 1324354657 ./MC8.105144.PythiaZee.py MC8.PythiaZee.110.evgen.pool.root

Submission through pathena works similarly:

> pathena --trf "csc_evgen08_trf.py <RunNo.> <FirstEvent> <MaxEvents> <RandomSeed> ./<JobOptions.py> %OUT.evgen.pool.root" 
       --outDS <OutputDataset> --noBuild

and our example requires little change:

> pathena --trf "csc_evgen08_trf.py 105144 1 10000 1324354657 ./MC8.105144 .PythiaZee.py %OUT.evgen.pool.root" 
       --outDS user09.MartinZeman.MC8.105144.PythiaZee.evgen.v14000502 --noBuild

Dont forget to change the name of the OutputDataset to user09.. otherwise pathena will not submit the job.

2.2 Simulation

Simulation is run just as any job in Athena using the athena.py script on a Job Options file. You can run simulation together with digitization using Geant4 by running csc_simul_trf.py script (accessible after sourcing Athena).

2.2.1 Running Simulation using CSC Job Transformation on Pathena

Locally you can run Simul using Job Transformation by issuing:

> csc_simul_trf.py <inputEvgenFile> <outputHitsFile> <outputRDOFile> <maxEvents> <skipEvents> 
       <randomSeed> <geometryVersion> <digiSeedOffset1> <digiSeedOffset2>

Submission through pathena:

> pathena --trf "csc_simul_trf.py <inputEvgenFile> <outputHitsFile> <outputRDOFile> <maxEvents> <skipEvents> 
       <randomSeed> <geometryVersion> <digiSeedOffset1> <digiSeedOffset2>" 
       --inDS <InputDataset>
       --outDS <OutputDataset>
       --noBuild --long

Following with the example:

> pathena --trf "csc_simul_trf.py %IN %OUT.hits.pool.root %OUT.rdo.pool.root 100 0 1324354656 ATLAS-CSC-02-00-00 100 1000" 
       --inDS user09.MartinZeman.MC8.105144.PythiaZee.evgen.v14000502
       --outDS user09.MartinZeman.MC8.105144.PythiaZee.simul.v14000502
       --noBuild --long

2.2.2 Running Athena on JobOptions?

This does not seem to work very well so I would encourage using the JobTransformation? . General guidelines here: https://twiki.cern.ch/twiki/bin/view/Atlas/PandaAthena#example_2_g4sim.

2.3 Digitization

Digitization is run together with simulation if using csc_simul_trf.py job transformation. You can use csc_digi_trf.py to run just digitization in the very same way.

2.4 Reconstruction

Reconstruction generally runs on the Reconstruction/RecExample/RecExCommon package. More information about how it works and ho write your JobOptions? can be found here: https://twiki.cern.ch/twiki/bin/view/Atlas/RunningReconstruction

Documentation: https://twiki.cern.ch/twiki/bin/view/Atlas/ReconstructionDocumentation

2.4.1 Running Reconstruction using CSC Job Transformation on Pathena

You can run reconstruction using Job Transformation as follows:

> csc_reco_trf.py <InputRDOFile> esd.pool.root aod.pool.root ntuple.root <MaxEvents> <Skip> 
       <GeometryVersion> <TriggerSettings>

which in pathena is:

> pathena --trf "csc_reco_trf.py <outputESDFile> <outputAODFile> <outputCBNTFile> <MaxEvents> <SkipEvents> 
       <GeometryVersion> <TriggerSettings>" 
       --inDS <InputDataset>
       --outDS <OutputDataset>
       --noBuild

The example:

> pathena --trf "csc_reco_trf.py %IN %OUT.ESD.pool.root %OUT.AOD.pool.root %OUT.CBNT.pool.root 100 0
       ATLAS-CSC-02-00-00 NONE" 
       --inDS valid1.005200.T1_McAtNlo_Jimmy.digit.RDO.e322_s405/ 
       --outDS user09.MartinZeman.pathenafiles.CSC.005200.Ideal.valid1.T1_McAtNlo_Jimmy.recon 
       --noBuild

Using non-CSC reconstruction Job Trasnformations does not seem to be supported by pathena.

3 Running multiple jobs

How this works has not yet been discovered.

--split
--nEventsPerJob
--nFiles

-- MartinZeman - 17 Feb 2009