Mass production of ntuples on data and (min bias) MC

Release

The working releases will change over time. The following have been tested to work.

• CMSSW_3_6_1
  cvs co -r ursl-100726 HeavyFlavorAnalysis/Bs2MuMu
  cvs co -r ursl-100726 AnalysisDataFormats/HeavyFlavorObjects
  

Datasets

Determine the dataset (for a given run) that you want to run over:

• Note:
  • dbs is available after cmsenv
  • More details on datasets and skims and versions can be obtained from Collisions2010Analysis

     dbs search --query='find dataset where run=136902'

List of files and events for each dataset

• Note:
  • You MUST do this on t3ui01, because dbs on t3ui02 does NOT know about still open blocks...

     dbs search --query='find file,file.numevents where dataset=/Mu/Run2010A-PromptReco-v*/RECO and site=t3se01.psi.ch and file.numevents>0' > file1

Create py files for n events, split into runs

• Note:
  • mkPyFiles is in Bs2MuMu/perl

     mkPyFiles -f file1 -t dimuons-XXXX.py -r -e 200000 -s r00

Create a tar file

This tarfile contains all necessary source code (for grid and/or batch submision)

• Note:
  • mkTarFile is in Bs2MuMu/perl
  • mkTarFile by default only includes the required ADF/HFO and HFA/Bs2MuMu packages. Additional packages can be added with -i if required.

         cd .../CMSSW_3_X_Y/src
         mkTarFile grid-100426.tar.gz
         mkTarFile [-i RecoVertex/KinematicFitPrimitives] grid-100426.tar.gz
    

Run the jobs

• Note:
  • run is in Bs2MuMu/perl
  • Of course, the relative paths need to be adapted to where you are running the production
  • Make sure that the directory referred to by STORAGE1 exists.
  • You can use (at your own risk) the perl skripts monSge (and monGrid) to monitor the progress of the jobs

     run -t ../../../../../../grid-100426.tar.gz -m batch -c ../../../test-local/ana.csh \
     -r 'STORAGE1 srm://t3se01.psi.ch:8443/srm/managerv2\?SFN=/pnfs/psi.ch/cms/trivcat/store/user/ursl/root/dana/v06-data'  \
     dimuons-r00-*.py

-- UrsLangenegger - 2010-07-26