Confusion in sequence when neither "--data" nor "--mc" is used

[srimanob]

We started to spot the issue when neither "--data" nor "--mc" is used. CMSSW normally guess using
https://github.com/srimanob/cmssw/blob/master/Configuration/Applications/python/cmsDriverOptions.py#L214-L235

The issue happened when it guesses as a data workflow, however, nothing (e.g. isData) is set after that. When we make a workflow with, i.e.
cmsDriver.py RECO --conditions 106X_dataRun2_v35 --customise Configuration/DataProcessing/Utils.addMonitoring --datatier NANOAOD --era Run2_2018,run2_nanoAOD_106Xv2 --eventcontent NANOEDMAOD --filein placeholder.root --fileout file:ReReco-Run2018C-JetHT-UL2018_MiniAODv2_NanoAODv9_pilot-00001.root --nThreads 2 --no_exec --python_filename ReReco-Run2018C-JetHT-UL2018_MiniAODv2_NanoAODv9_pilot-00001_0_cfg.py --scenario pp --step NANO
it will guess it is data workflow, but pick up MC customization. The issue is not only on NanoAOD, Mini will be the same.

The PR #34302 introduces a fix.

However, when we start to set isData properly, we start to see a failure in IB test. The issue come from DQM validation sequence, as reported in #34438
In short, the validation sequence try to run
cmsDriver.py step3 --conditions auto:run2_data -s VALIDATION --process DUMMY --eventcontent DQM --datatier DQMIO --customise_commands 'process.Tracer = cms.Service("Tracer")' --filein file:/eos/cms/store/data/Run2018A/EGamma/RAW/v1/000/315/489/00000/004D960A-EA4C-E811-A908-FA163ED1F481.root -n 10 --python_filename cmssw_cfg.py --no_exec
which is data workflow. However, MC sequence is picked including pileup replay. Config can be dumped, but it does not make sense. The dump step will fail when "--data" or "isData" is used.

It seems the test of DQM sequence should be reviewed first to avoid broken in data workflow. Then we can apply #34302 again, to make mini/nano data workflow properly when it comes from CMSSW guess.

[cmsbuild]

A new Issue was created by @srimanob Phat Srimanobhas.

@Dr15Jones, @perrotta, @dpiparo, @silviodonato, @smuzaffar, @makortel, @qliphy can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[qliphy]

assign dqm,xpog

[cmsbuild]

New categories assigned: dqm,xpog

@jfernan2,@kmaeshima,@fgolf,@mariadalfonso,@rvenditti,@andrius-k,@gouskos,@ErnestaP,@ahmad3213 you have been requested to review this Pull request/Issue and eventually sign? Thanks

[jfernan2]

@srimanob I have been looking into this but my knowledge of the PdmV configs from cmsDriver is limited.
Nevertheless, enabling your PR #34302 and doing a print before the processing of the 10 sequences crashing in:
https://github.com/cms-sw/cmssw/blob/master/DQMOffline/Configuration/scripts/cmsswSequenceInfo.py#L530

after scram b runtests_TestDQMOfflineConfiguration_110

I get:

Config file cmssw_cfg.py created
[Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Phase2C11I13T25M9', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Phase2C11I13T26M9', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Phase2C9', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016', scenario='pp', mc=True, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016,run2_miniAOD_80XLegacy', scenario='pp', mc=True, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016_HIPM', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016_HIPM', scenario='pp', mc=True, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2016_trackingLowPU', scenario='', mc=False, data=False, fast=False), Sequence(seqname='@miniAODValidation', step='VALIDATION', era='Run2_2017', scenario='', mc=False, data=False, fast=False)]
Analyzing 10 seqs...

So, it looks like there are several sequences with mc=False and data=False at the same time...Is this expected?

[srimanob]

Hi @jfernan2
I don't have an idea why it is not set properly. Do you know how these workflows have been selected?

However, the clues I have include

1. The input and GT used in DQM sequence validation is just a dummy. Since we don't work on physics result with it.
2. With validation sequence only, CMSSW will guess it is a data workflow. However, it will pick up MC sequence because isData is not set after the guess.
3. miniaodValidation will not work with data. Current things work because of a bug in (2).

The fix I propose in #34452 include

1. Assign properly "isData" when it is data workflow
2. Removing mixing replay when the workflow is data.
3. Add miniaodValdiation and Phase2 as a "temporary" check for MC. This I can disable when I know how to access the test workflow of DQM, and define properly what the workflow should be.

[jfernan2]

Your solution seems fine to me, thanks!

I would not consider them selected workflows but module/sequences coming from a full runTheMatrix inspection. They are part of the UnitTests under DQMOffline/Configuration/test which just grabs any DQM,Validation module/sequence to check its validity:

https://github.com/cms-sw/cmssw/blob/master/DQMOffline/Configuration/test/runtests.sh#L23

Legacy modules are stripped and then the full list of sequences is splitted in chuncks to run them
https://github.com/cms-sw/cmssw/blob/master/DQMOffline/Configuration/test/BuildFile.xml

[srimanob]

@cms-sw/dqm-l2

Could you please confirm the statement in
#34452 (comment) ?

For step:

• VALIDATION => MC
• VALIDATION + DQM => MC
• DQM => Data or MC. In this case, running in Data mode with MC file should be OK.
  ?

Thanks in advance.

[jfernan2]

+1
Solved by #3445

[mariadalfonso]

type jetmet

[mariadalfonso]

This explicit --data and --mc is only needed by the MET sequence.

Suggest to review the METcode such that like ANY other object doesn't rely on external switches

[vlimant]

+1
we took note of the MET configuration that needs cleaning

[cmsbuild]

This issue is fully signed and ready to be closed.