Should data tier of EXODisappTrk skim be AOD?

[makortel]

While investigating https://cms-talk.web.cern.ch/t/getting-error-readeventsucceeded-failed-when-running-cmsrun/29893/1 it was noticed that the EXODisappTrk skim (e.g. this dataset /Muon/Run2022D-EXODisappTrk-27Jun2023-v2/AOD uses AOD as data tier, but the event content is very different from AOD (last entries of the outputCommands of the relevant output module

        'drop *',
        'keep *_reducedHcalRecHits_*_*',
        'keep *_reducedEcalRecHits*_*_*'

). Using AOD data tier here sounds potentially confusing.

[makortel]

assign pdmv

[cmsbuild]

New categories assigned: pdmv

@AdrianoDee,@sunilUIET,@miquork you have been requested to review this Pull request/Issue and eventually sign? Thanks

[cmsbuild]

A new Issue was created by @makortel Matti Kortelainen.

@antoniovilela, @makortel, @smuzaffar, @rappoccio, @sextonkennedy, @Dr15Jones can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[mmusich]

looks like this happens because it first clones AODSIM, then removes everything and just adds the reduced E/HCAL rechtis?

cmssw/Configuration/Skimming/python/PDWG_EXODisappTrk_cff.py

Lines 3 to 7 in 56711f4

	from Configuration.EventContent.EventContent_cff import AODSIMEventContent
	EXODisappTrkSkimContent = AODSIMEventContent.clone()
	EXODisappTrkSkimContent.outputCommands.append('drop *')
	EXODisappTrkSkimContent.outputCommands.append('keep *_reducedHcalRecHits_*_*')
	EXODisappTrkSkimContent.outputCommands.append('keep *_reducedEcalRecHits*_*_*')

[makortel]

@cms-sw/pdmv-l2 Any thoughts?

[youyingli]

The EXODisappTrk skim only stores reduced RecHits for ECAL and HCAL which belong to AOD content (here). The first PR was created in #36846 and the AOD data tiar was required in this skim (from here), perhaps because reduced RecHits belong to AOD content. @carriganm95, could you please explain why the skim needs AOD data tiar instead of USER let's say?

[carriganm95]

@youyingli I am not sure I understand the nuances here but I can explain what we want out of this skim. Our analysis needs these AOD rec hits collections and otherwise uses miniAOD data. In the lead up to run 3 the EXO group mentioned that anyone who wants AOD content will need to make a skim of the collections they want because AOD will not be available to users. So we created this skim as a centrally produced skim that we can use to get our rec hits. If the data tier is USER will it still be centrally produced?

[youyingli]

@carriganm95 Thanks for the explanation. I think @makortel would like to understand why this skim is labeled to AOD shown in DAS as the skim drops all AOD contents but keeps only ECAL and HCAL reduced recHits. To change here to USER is still centrally produced. @makortel you think changing from AOD to USER is more sultable?

[carriganm95]

@youyingli One other thing that may be relevant and I forgot to mention is that we run by using the AOD and miniAOD as sibling files in our crab jobs. So it is necessary for us that the ancestry continues to allow this, I am not sure how the ancestry would work for USER files.

[davidlange6]

The name of the data tier has no impact on this On Oct 13, 2023 6:33 AM, carriganm95 ***@***.***> wrote: @youyingli<https://github.com/youyingli> One other thing that may be relevant and I forgot to mention is that we run by using the AOD and miniAOD as sibling files in our crab jobs. So it is necessary for us that the ancestry continues to allow this, I am not sure how the ancestry would work for USER files. — Reply to this email directly, view it on GitHub<#42893 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABGPFQ45S5KXZNYRCGOJSI3X7C77ZANCNFSM6AAAAAA5JT432E>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[makortel]

One other thing that may be relevant and I forgot to mention is that we run by using the AOD and miniAOD as sibling files in our crab jobs. So it is necessary for us that the ancestry continues to allow this,

Note that this setup (using sibling files with two-file solution) is NOT formally supported (see https://cms-talk.web.cern.ch/t/getting-error-readeventsucceeded-failed-when-running-cmsrun/29893/9).

[makortel]

you think changing from AOD to USER is more sultable?

I think it would reduce confusion by making it clear that the event content is not AOD. (but I'm still somewhat confused of the formal distinction of "data tier" and "event content")