This is an example CRAB config to submit locally to Vanderbilt using PBS instead of glidein. It is VERY important that you not have more than 5000 queued jobs in the cluster. Please use the CMSSW and CRAB installations in CVMFS:
. /cvmfs/cms.cern.ch/cmsset_default.sh . /cvmfs/cms.cern.ch/crab/crab.sh
You then can use/modify the following config to submit.
[CRAB] jobtype=cmssw scheduler=pbsv2 use_server=0 [CMSSW] datasetpath=/unweighetedEvents2/meloam-unweighetedEvents2-2a0c99fbde35e8a44e0fd856790266bd/USER dbs_url=http://cmsdbsprod.cern.ch/cms_dbs_ph_analysis_02/servlet/DBSServlet pset=unweighted_RAW.py events_per_job=1 number_of_jobs=10 get_edm_output=1 [USER] check_user_remote_dir=0 copy_data=1 return_data=0 storage_element=T2_US_Vanderbilt publish_data=1 publish_data_name=unweighetedEventsRaw2test ui_working_dir=unweightedEventsRaw2test dbs_url_for_publication=https://cmsdbsprod.cern.ch:8443/cms_dbs_ph_analysis_02_writer/servlet/DBSServlet [GRID] se_white_list=T2_US_Vanderbilt [PBSV2] grouplist=jswhep workernodebase=/tmp hostname=vmps09.vampire # YOU MUST TUNE THE WALL TIME TO THE CORRECT LENGTH # Too long a wall time will make your jobs less likely to get scheduled # Too short a wall time will cause your jobs to get slayed by the scheduler resources=walltime=10:00:00,pmem=2800mb,mem=2800mb