Remove ESProducers for ECAL Mustache parameters from configuration

[thomreis]

PR description:

Since the records should now be provided with the GT this PR removes the ESSource and ESProducers for EcalMustacheSCParameters and EcalSCDynamicDPhiParameters from the default configuration.

PR validation:

Passes the limited matrix tests.

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-33183/21594

• This PR adds an extra 12KB to repository

[perrotta]

please test

[cmsbuild]

A new Pull Request was created by @thomreis (Thomas Reis) for master.

It involves the following packages:

RecoEcal/EgammaClusterProducers

@perrotta, @jpata, @cmsbuild, @slava77 can you please review it and eventually sign? Thanks.
@Sam-Harper, @lecriste, @rchatter, @jainshilpi, @rovere, @lgray, @sobhatta, @thomreis, @afiqaize, @simonepigazzini, @argiro, @varuns23, @ram1123 this is something you requested to watch as well.
@silviodonato, @dpiparo, @qliphy you are the release manager for this.

cms-bot commands are listed here

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-73752a/13525/summary.html
COMMIT: 4bdc602
CMSSW: CMSSW_11_3_X_2021-03-14-2300/slc7_amd64_gcc900
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/33183/13525/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 14470 differences found in the comparisons
• DQMHistoTests: Total files compared: 37
• DQMHistoTests: Total histograms compared: 2635087
• DQMHistoTests: Total failures: 9185
• DQMHistoTests: Total nulls: 6
• DQMHistoTests: Total successes: 2625874
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 36 files compared)
• DQMHistoSizes: changed ( 1000.0,... ): 0.004 KiB JetMET/SUSYDQM
• DQMHistoSizes: changed ( 136.731 ): -0.004 KiB JetMET/SUSYDQM
• DQMHistoSizes: changed ( 312.0 ): -0.004 KiB MessageLogger/Warnings
• Checked 155 log files, 37 edm output root files, 37 DQM output files

[perrotta]

• Reco comparison results: 14470 differences found in the comparisons

It doesn't seem they were so unused...

[thomreis]

Not unused but fixed to one specific set of parameters. The ESProducer produced parameters for Run 3 for all WFs. Now with the records coming from the GT there are also the Run 2 parameters used for certain WFs. There I would expect some differences. I need to verify though that the WFs with differences use tags with Run 2 parameters.

[thomreis]

I have checked all WFs with failing comparisons and they all use tags with Run 2 parameters, apart from WF 312.0, which has only failed comparisons in the MessageLogger.

[perrotta]

+1

• It removes the ESSource and ESProducers for EcalMustacheSCParameters and EcalSCDynamicDPhiParameters, so that those parameters are now taken from the official tag in the DB
• With this change, those run2 (and run1) parameters are used in run2 (and run1) workflows, and not the run3 ones which were used before in all cases: this originates the differences visible in the outputs of the jenkins tests for all workflows previous to run3

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @silviodonato, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2)

[silviodonato]

+1
related to #32997 and #32959

[slava77]

Now with the records coming from the GT there are also the Run 2 parameters used for certain WFs. There I would expect some differences.

this was discussed in the ORP earlier today.
The comparisons show that the changes are not from numerical precision. So, the change is significant.
Is there an intended difference between the GT and the hardcoded parameters for Run 2?

@christopheralanwest

[thomreis]

The hardcoded parameters that the ESProducer used were the ones for Run 3 for all WFs. Now with the parameters coming from the GT the parameters can be either the Run 2 or the Run 3 ones depending on the WF. All WFs that show differences get the Run 2 parameters from the GT now instead of the previously hardcoded Run 3 ones.

[christopheralanwest]

The hardcoded parameters that the ESProducer used were the ones for Run 3 for all WFs. Now with the parameters coming from the GT the parameters can be either the Run 2 or the Run 3 ones depending on the WF. All WFs that show differences get the Run 2 parameters from the GT now instead of the previously hardcoded Run 3 ones.

If this is the case, why were there no comparison test differences in the (non-DD4HEP) comparison tests for PR #32066 (see #32066 (comment))? The absence of any differences when these ESProducers were introduced is what convinced me that this was a purely technical change.

[thomreis]

Because the change from hardcoded Run 2 parameters to hardcoded Run 3 parameters was already done before #32066 with PR #31823. And there are differences observed and discussed in that PR.

[christopheralanwest]

So is it fair to say that PR #31823 renders Run 2 workflows sub-optimal by an amount that was seen as acceptable at the time but this PR restores the original behavior?

[thomreis]

I think that can be said for Run 2 WFs, yes.