LogError in HLTPrescaleProvider used by PATTriggerProducer:patTrigger: Error in determining L1T prescale

[slava77]

this can be seen in workflow 136.78 step3:

%MSG-e HLTPrescaleProvider:  PATTriggerProducer:patTrigger  18-Jul-2017 18:52:46 CEST Run: 297227 Event: 32100
 Error in determining L1T prescale for HLT path: 'HLT_ZeroBias_FirstCollisionInTrain_v2' has multiple L1TSeed modules, 2, with L1T seeds: 'L1_FirstCollisionInTrain' * 'NOT L1_FirstCollisionInOrbit'. (Note: at most one L1TSeed module is allowed for a proper determination of the L1T prescale!)
%MSG-e HLTPrescaleProvider:  PATTriggerProducer:patTrigger  18-Jul-2017 18:52:46 CEST Run: 297227 Event: 32100
 Error in determining L1T prescale for HLT path: 'HLT_ZeroBias_LastCollisionInTrain_v1' has multiple L1TSeed modules, 2, with L1T seeds: 'L1_FirstCollisionInTrain' * 'L1_LastCollisionInTrain'. (Note: at most one L1TSeed module is allowed for a proper determination of the L1T prescale!)

@Martin-Grunewald @fwyzard @gpetruc @arizzi

[slava77]

assign hlt,reconstruction

[cmsbuild]

A new Issue was created by @slava77 Slava Krutelyov.

@davidlange6, @Dr15Jones, @smuzaffar can you please review it and eventually sign/assign? Thanks.

cms-bot commands are listed here

[cmsbuild]

New categories assigned: hlt,reconstruction

@Martin-Grunewald,@slava77,@perrotta,@silviodonato,@fwyzard you have been requested to review this Pull request/Issue and eventually sign? Thanks

[Martin-Grunewald]

Hi - indeed I see this, eg, in 93X.

These messages are limited to at most 5 per instance of the HLTPrescaleProvider.

One of the two paths, namely HLT_ZeroBias_LastCollisionInTrain, is fixed with the new HLT menu integrated in 93X yesterday, which avoids two HLTL1T seed modules in the path.

I need to check why this construct was used by the remaining trigger HLT_ZeroBias_FirstCollisionInTrain - discussed here: https://its.cern.ch/jira/browse/CMSHLT-1281

[Martin-Grunewald]

The remaining trigger is fixed and will become available with HLT menu V2.1.

[slava77]

@Martin-Grunewald
I notice that the "Error in determining L1T prescale" error is still showing up even in 2018 data processing.
e.g. in 10_3_0_pre5 in wf 136.85 Run2018A EGamma PD

%MSG-e HLTPrescaleProvider:  PATTriggerProducer:patTrigger  06-Oct-2018 21:53:46 CEST Run: 315489 Event: 47697292
 Error in determining L1T prescales for HLT path: 'HLT_DiEle27_WPTightCaloOnly_L1DoubleEG_v4' with complex L1T seed: [long list of seeds omitted]

[Martin-Grunewald]

@slava77
This is a valid error message which goes off after 5 instances or so.
PATTriger would have to be fixed...

[missirol]

Trying to tackle this old issue.

This type of error can still be seen in recent RelVals, e.g.
https://cmssdt.cern.ch/SDT/jenkins-artifacts/ib-baseline-tests/CMSSW_13_0_X_2023-01-10-2300/el8_amd64_gcc11/-GenuineIntel/matrix-results/136.8311_RunJetHT2017FreMINIAOD/step2_RunJetHT2017FreMINIAOD.log

Short version : I suspect the issue is the choice of GT (no HLT/RECO/PAT changes needed); the problematic steps use auto:run*_data_relval while auto:run*_data should be used instead.

Longer version : I think the behavior of HLTPrescaleProvider is correct (it reports that expected L1T seeds are not found; one could argue on whether that should be a LogWarning or a LogError). This relates to the L1T menu used by the job, which in turn goes back to the GT. In the example above, the GT is auto:run2_data_relval. The latter is the same as auto:run2_data except that in auto:run2_data_relval the L1T menu tag is a single-IOV tag with a specific L1T menu (in principle, the 'latest Run-2 L1T menu in use for this release cycle'); in this case, this means that this 2017 wf loads the 2018 L1T menu, but reads the HLT results from the 2017 EDM file (2017 HLT menu), thus the mismatch. Loading the "latest" L1T menu (i.e. using auto:run*_data_relval) would make sense if the wf also included a step with "L1REPACK + (re)HLT".

I guess that, as a tentative rule of thumb, if the RelVal step (1) contains --data and (2) does not contain --(p|hltP)rocess reHLT, then its GT should not be a auto:*_relval one [*].

[*] Some steps with -s HARVESTING:* --filetype DQM or -s SKIM:LogError+LogErrorMonitor fall in this category (i.e. use --conditions auto:*_data_relval without --(p|hltP)rocess reHLT), and they might be considered exceptions to this rule-of-thumb (these cases are probably harmless; I doubt they consume the L1T menu in the GT, but I don't really know).

[missirol]

+hlt