Submit script with crab

This page demonstrates the use of the CMS Remote Analysis Builder (CRAB) with a custom script instead of cmsRun

CRAB 2

A simple example

Environment

First, lets set up the environment

cmsrel CMSSW_5_3_18
cd CMSSW_5_3_18/src
cmsenv
source /cvmfs/cms.cern.ch/crab/crab.sh

Script to be run

Next we need the script we want to run. This example will print 'Hello CRAB!' on the screen and create a ROOT file named 'testing.root'.

echo "# write something to the console
echo "Hello CRAB!"
# create a dummy file
touch testing.root" > run.sh
chmod a+x run.sh

The CRAB config

Finally, we need to create a CRAB config in order to submit our script to the grid. The example below tells CRAB which script to run and what output files to expect.

echo "[CRAB]
jobtype = cmssw
scheduler = remoteGlidein 
use_server = 0

[CMSSW]
output_file = testing.root
datasetpath=None
pset=None
number_of_jobs=1
allow_NonProductionCMSSW = 1

[USER]
script_exe = run.sh 
ui_working_dir = dummy
copy_data = 0
return_data = 1" > crab.cfg

Submission to the grid

Everything should be now ready to submit our script to the grid

crab -create -cfg crab.cfg
crab -submit -c dummy

Now it is time to wait. You can check the status of your job via

crab -status -c dummy

Once the job is done, it is time to get the output

crab -get -c dummy

should put all the output files into dummy/res. The standard output should contain our echo:

cat dummy/res/CMSSW_1.stdout  | grep CRAB
env var CRAB_UNIQUE_JOB_ID set to: kreczko_dummy_7qp08n_1_1_Yhj
# LIMITS USED BY CRAB WATCHDOG:
Hello CRAB!
CRAB FrameworkJobReport crab_fjr.xml is not available, using exit code of executable from command line.
# LIMITS USED BY CRAB WATCHDOG:

and our test ROOT file should be in this folder as well

 ls -lah dummy/res/
total 32K
drwxr-xr-x 2 phxlk users 4.0K Jul 13 16:38 .
drwxr-xr-x 6 phxlk users 4.0K Jul 13 16:24 ..
-rwx------ 1 phxlk users    0 Jul 13 16:26 CMSSW_1.stderr
-rwx---r-- 1 phxlk users  16K Jul 13 16:36 CMSSW_1.stdout
-rw------- 1 phxlk users  804 Jul 13 16:33 Watchdog_1.log.gz
-rw------- 1 phxlk users  568 Jul 13 16:36 crab_fjr_1.xml
-rw------- 1 phxlk users    0 Jul 13 16:26 testing_1_1_Yhj.root

If you see the above, everything worked as it should!

Advanced example

If you have a look at the script CRAB produces in the previous example you can see the following:

 cat dummy/job/CMSSW.sh | grep run.sh
mv $RUNTIME_AREA/run.sh . 
/usr/bin/time -f "%U %S %P" -o cpu_timing.txt $executable run.sh $NJob $AdditionalArgs > executable.out 2>&1

Our script has been called with the job number ($NJob) as the first parameter followed by zero or more additional parameters ($AdditionalArgs). These additional parameters can be set using a comma separated list under the [USER] section:

script_arguments= a,b,c

Inside the script the job number will be available as $1 and all following parameters as $2 etc.

CRAB 3

Not available (yet).