cmsrel CMSSW_10_2_11
cd CMSSW_10_2_11/src/
cmsenv

# clone this repo into "DeepNTuples" directory
git clone https://github.com/hqucms/DNNTuplesAK8.git DeepNTuples -b 94X
scram b -j24

Step 0: switch to the crab production directory and set up grid proxy, CRAB environment, etc.

cd $CMSSW_BASE/src/DeepNTuples/Ntupler/run
# set up grid proxy
voms-proxy-init -rfc -voms cms --valid 168:00
# set up CRAB env (must be done after cmsenv)
source /cvmfs/cms.cern.ch/crab3/crab.sh

Step 1: use the crab.py script to submit the CRAB jobs:

python crab.py --set-input-dataset -p ../test/DeepNtuplizerAK8.py --site T2_CH_CERN -o /store/user/$USER/DeepNtuples/[version] -t DeepNtuplesAK8-[version] --no-publication -i [ABC].conf -s FileBased -n 5 --work-area crab_projects_[ABC] --dryrun

These command will perform a "dryrun" to print out the CRAB configuration files. Please check everything is correct (e.g., the output path, version number, requested number of cores, etc.) before submitting the actual jobs. To actually submit the jobs to CRAB, just remove the --dryrun option at the end.

[Note] For the QCD samples use -n 1 --max-units 20 to run one file per job, and limit the total files per job to 20.

Step 2: check job status

The status of the CRAB jobs can be checked with:

./crab.py --status --work-area crab_projects_[ABC]

Note that this will also resubmit failed jobs automatically.

The crab dashboard can also be used to get a quick overview of the job status: https://dashb-cms-job.cern.ch/dashboard/templates/task-analysis

More options of this crab.py script can be found with:

./crab.py -h