/submission

Primary LanguagePython

Submission tools for batch job management (CERN HT-Condor & CMS CRAB)

How-to submit jobs

Inline help:

python3 submit.py --help

Create the Job configuration using the command:

python3 submit.py -f submit.yaml --create

Submit the jobs to the queues via the command:

python submit.py -f submit.yaml --submit

Configuration file

The conf file uses the YAML format. You can find several examples in this directory. Most of the parameters should be self explenatory. Each task corresponds to a sample.

The tool supports:

  • HT-Condor local submission: each task will be mapped to a dedicated condor cluster. Input files can be specified via the dataset name using input_dataset or via the directory where the files sit using input_directory. The implemented input splitting modes are: file_based and lumi_based (in the sense of LS). For the job_flavor and other condor parameters please refer to the CERN Condor documentation

  • CRAB remote submission: this is controlled setting the crab: True parameter. In this case the input needs to be specified using input_dataset and for the configuration of splitting_mode and splitting_granularity you need to refer to the CMS CRAB Sw Guide. The rest of the parameters can be customized via the template file crab_MODE.py (see the documentation about the template structure in the following).

How to modify the template structure

If you want to add a new set of templates you need to define a new mode (MODE) in:

https://github.com/cerminar/submission/blob/master/submit.py#L101

Note the code assumes that all keys starting with TEMPL_ are variables in the templated files that need to be replaced.

And create the corresponding template files in the directory templates/

You need 1 or 2 templated files:

  1. condorSubmit_MODE.sub: this defines the structure of the Condor submission file. If you don't need to modify it you can use the condorSubmit_DEFAULT.sub one (see CERN Condor documentation);
  2. jobCustomization_MODE_cfg.py. which defines what needs to be changed for each job of the cluster.
  3. run_MODE.sh: this is the script which actually gets executed on the node, setups the config and calls cmsRun. Again a run_DEFAULT.sh exists and should work for most cases.