Crab Grid Test

Goal

Investigate performance of glidein submission versus direct (condor) submission

Implementation

Run ntuplemaker on dataset

We chose the ntuplemaker because it is a useful thing to run, and nontrivial

• Check out CMSSW (we use 2_1_9 for this exercise, this will create a directory in the pwd where it is called)

scramv1 project CMSSW CMSSW_2_1_9

• Check out CMS2

export CVSROOT=:pserver:anonymous:98passwd@cmscvs.cern.ch:/cvs_server/repositories/CMSSW/
cd CMSSW_2_1_9/src
cvs co -r V01-00-04 -d CMS2/NtupleMaker UserCode/JRibnik/CMS2/NtupleMaker
scramv1 b -j 8

• Edit the CMSSW config file
  • There should be a file 'test_cfg.py' in 'CMSSW_2_1_9/src/CMS2/NtupleMaker/test'
  • Change the data to run on (optional?): line 91:

fileNames = cms.untracked.vstring('/store/relval/CMSSW_2_1_9/RelValZEE/GEN-SIM-DIGI-RAW-HLTDEBUG-RECO/STARTUP_V7_v2/0000/04419036-F385-DD11-B3A7-001617C3B6E8.root')

• • Change output (broken--we don't get any output of first test)
  • Change path to only run a single module instead of all defined: line 157 (second to last):

process.p = cms.Path(process.generalmakers)

• Run CMSSW: (note: the first line is common to alias to 'ev')

eval `scramv1 runtime -sh`
cmsRun test_cfg.py

Install Crab and Submit Job

• Install Crab

wget https://cmsweb.cern.ch/crabconf/files/CRAB_2_4_2.tgz --no-check-certificate
tar -zxvf CRAB_2_4_2.tgz
cd CRAB_2_4_2
./configure

• Prepare crab job
  • Go on DBS (link below) and find the dataset you put in your CMSSW config file. Under that dataset, you should see a link 'crabCfg'. This is a sample crab config file for this dataset.
  • Put this file in 'CMS2/NtupleMaker/test/test_crab.cfg'
  • Edit this file:
    • Change scheduler to 'condor_g' instead of 'glite' (line 4 or so)
    • Make sure 'datasetpath' is correct (line below '[CMSSW]')
    • Insert (edit) the following line immediately below the 'datasetpath' line to tell Crab which CMSSW config file to use:

pset=test_new.py

• • • Make sure there is no typo in the following line (at end of CMSSW section):

events_per_job = 50

• • • Optionally, change total events, or events per job if desired
    • If there is a problem with site selection (or if you want to be safe and be sure you run at UCSD), insert the following at the end of this file:

[EDG]
SE_white_list = srm-3.t2.ucsd.edu

• Source:

source ~/CRAB_2_4_2/crab.sh
source /code/osgcode/ucsdt2/gLite/etc/profile.d/grid_env.sh

• Submit crab job
  • In order for this to work, you will need a proxy, or a grid cert which crab will use to get you a proxy
  • 2 options:
    • To create and submit the job in one command:

crab -create -submit -cfg test_crab.cfg

• • • Optionally, you can create the job and submit separately (where xxx is last output from the first line below):

crab -create -cfg test_crab.cfg
crab -submit -continue xxx

• Check status of job (xxx is optional, and defined line above):

crab -status -continue xxx

• Get output of job: is dir created by crab, optional: is given by the status command above, this should work for completed jobs

crab -getoutput -c

Links

https://cmsweb.cern.ch/dbs_discovery/

http://hepuser.ucsd.edu/twiki2/bin/view/UCSDTier2/PhysicsAndMCproduction

https://cmsweb.cern.ch/crabconf/files/crab-v2.4.2.html

-- WarrenAndrews - 2008/12/01