CRAB recipe for PF validation about cmssw

[hatakeyamak]

Thanks.

I am not sure what "CMSSW_10_5_0_pre1 is not available" means though.

ls $CMS_PATH/slc*/cms/cmssw | grep CMSSW_10_5
CMSSW_10_5_0_pre1/
CMSSW_10_5_0_pre2/
CMSSW_10_5_0_pre1/
CMSSW_10_5_0_pre1/
CMSSW_10_5_0_pre1_ROOT614/
CMSSW_10_5_0_pre2/
CMSSW_10_5_0_pre2_G4v1043/
CMSSW_10_5_0_pre1/
CMSSW_10_5_0_pre2/

so these releases are available on cvmfs, and crab jobs can use them. It looks to me that pre releases are just not on the file you link, but it doesn't mean the release is not available I think.

[jpata]

It seems there is a difference between being available on cvmfs and usable by crab. You get errors when trying to submit on releases not on the list above.

…

On 26 Feb 2019, at 17:50, Kenichi Hatakeyama ***@***.***> wrote: Thanks. I am not sure what CMSSW_10_5_0_pre1 means though. ls $CMS_PATH/slc*/cms/cmssw | grep CMSSW_10_5 CMSSW_10_5_0_pre1/ CMSSW_10_5_0_pre2/ CMSSW_10_5_0_pre1/ CMSSW_10_5_0_pre1/ CMSSW_10_5_0_pre1_ROOT614/ CMSSW_10_5_0_pre2/ CMSSW_10_5_0_pre2_G4v1043/ CMSSW_10_5_0_pre1/ CMSSW_10_5_0_pre2/ so these releases are available on cvmfs, and we can use them. It looks to me that pre releases are just not on the file you link, but it doesn't mean the release is not available I think. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[hatakeyamak]

Hmm. What kind of error? I do crab with pre releases all the time I believe.

[jpata]

By default, crab just refuses to submit. It suggests to use config.JobType.allowUndistributedCMSSW = True if "you know what you're doing".

Earlier on, I was getting some strange failures from the jobs (can't recall any more) when I did set this flag, which I was attributing to this undistributed issue. Now trying again and being more careful with maxMemoryMB etc, the jobs seem to be running, so lets see if they succeed.

[hatakeyamak]

ok. sounds good.

[jpata]

Crab works, the recipe is in crab/multicrab.py. I was able to run all 4 datasets with no errors in a few hours. Merged in PR #34.

[hatakeyamak]

Excellent!