Closed belforte closed 1 year ago
Yes, I think that I could benefit from being involved, thanks Stefano!
First step: I replicated the behavior.
220727_124006:dmapelli_crab_20220727_144002
[1]skipEvents = 290
, processed 10 events: 220727_151454:dmapelli_crab_20220727_171452
[2]skipEvents = 0
, processed 300 events: 220727_153633:dmapelli_crab_20220727_173631
[3]After some time wondering why I did not see the effects of my changes to TweakPSet.py
, I realized I should
./updateTMRuntime.sh
I submitted a new task 220728_140056:dmapelli_crab_20220728_160054
and it looks good [1]!
[1] I will add all the details in the PR that I am about to open.
of course you need to ./updateTMRuntime.sh
before you start the TW process :-)
with reference to https://cms-talk.web.cern.ch/t/partial-processing-with-automatic-splitting-in-crab-job/13093/12 we found that CRAB wrapper, at PSet tweaking time, does not make sure that skipEvents is set to zero. In case the input PSet has a non-zero value there, it would be propagated and mess up processing. CRAB needs to make sure that all lumis indicated in input to a file will be processed.
IN this are it is more important to be careful than to push a quick fix, all in all we have been running like this since ever, as far as I can see.
@mapellidario let me know if you'd like to be involved as 'getting to know the details of job wrapper"