Report on PdmV Activities G. Boudoul, G. Franzoni, D. Piparo, J-R - - PowerPoint PPT Presentation

Report on PdmV Activities

• G. Boudoul, G. Franzoni, D. Piparo, J-R Vlimant

In a Nutshell

• Validation for 6.1 has to catch up and continue.
• ValDB is now in full production.
• For HLT validation, we need focus on DPG level

validation, and more attention on Data.

• Validation for condition update @HLT and @Prompt

run smoothly.

• Planning for Upgrade Validation coming are coming

together.

• Event different in different dataset understood, and

the magnitude of issue is being investigated.

• HCal laser pollution issue: the full list for all Pds is

coming together. There already exists datasets that can be scrutinized.

PPD Meeting, PdmV report, J-R Vlimant

Validation Organization

• Provide links to releases notes

➢ Maintained by the relevant experts

• Provide links to validation tasks:

➢ Offline, Conditions, HLT, Physics

• List of contacts

✔ https://twiki.cern.ch/twiki/bin/view/CMS/PdmV#Contacts ➢ Solely maintained in the relevant e-groups : no list maintained in twikis anymore ➔ Please notify PdmV conveners of any change, so as to keep efficient communications

• Validation Tables

✔ https://twiki.cern.ch/twiki/bin/view/CMS/PdmVValidationTables61X ➢ Deprecated and replaced by the usage of valDB

https://twiki.cern.ch/twiki/bin/view/CMS/PdmVValDB

➢ Many issues fixed already, please report to us any problem or suggestion you might have

➔ From 6.1 pre4 on, only filing with valDB is necessary  HN post to relevant group is handled by valDB  Summary view in valDB ➔ (planning on twiki-like representation)

https://twiki.cern.ch/twiki/bin/view/CMS/PdmVTasks#Validation

PPD Meeting, PdmV report, J-R Vlimant

valDB

https://twiki.cern.ch/twiki/bin/view/CMS/PdmVValDB Many thanks to Antanas

PPD Meeting, PdmV report, J-R Vlimant

6.1 Validation Status

✔ 6.1 validation in valDB : http://tinyurl.com/8ljuzhs ➔ Keep up with the good work of validation, for those filing reports ➔ Coverage is still scarce and regardless of the scope of 6.1,

validation must be performed.

PPD Meeting, PdmV report, J-R Vlimant

Condition Update Validation

• Validation of weekly Cond@HLT

✔ https://twiki.cern.ch/twiki/bin/view/CMS/PdmVTriggerConditionValidation ➢ Has been running smoothly weekly by TPG  ReRun HLT with and w/o new conditions on an recent run  Restricted to hand full of validators in feedback ➔ In smooth production mode again

• Validation of bi-weekly Cond@PromptReco

✔ https://twiki.cern.ch/twiki/bin/view/CMS/PdmVPRConditionsValidation ➢ Has been running smoothly bi-weekly by PdmV  ReRun RECO with and w/o new conditions on a recent run  Restricted to hand full of validators in feedback  relMon between two different runs : available ➔ Keep tight connection with validators for quickest feedback. ➢ Planning is bi-weekly, any more frequent will have to make a case to

PPD Coordination

➔ More in Loic's presentation today  Planning for full automation of the procedure down to validator feedback ➔ Using valDB ➔ Requires gui/web-interface design, DQM & valDB api

PPD Meeting, PdmV report, J-R Vlimant

6.0 / Upgrade Validation

• Effort to integrating release validation on upgrade release on-going:

➔ Workflows integrated in standard matrix test ➔ New manpower for development, please volunteer to DPG/POG/PAG/TSG

• Release Validation Production

➢ Set of upgrade /GEN-SIM sample made in 6.0.1 ➔ More in presentations from Piet and Vadim today. ➢ “Identity test” for current geometry ✔ Exact identity test failed in production (understood, fixed, but not in time) ➢ “Almost idenity test” performed with 6.0.0 vs 6.0.1



http://cms-service-reldqm.web.cern.ch/cms-service-reldqm/ReleaseMonitoringSpecial/601_Full/



http://cms-service-reldqm.web.cern.ch/cms-service-reldqm/ReleaseMonitoringSpecial/601_Fast/

➔ No relevant differences observed: Green light

• Plans

➢ Add the DQM developers and experts in the MC validation e-group to keep the

communication with validators tight.

➢ Use of valDB ➢ Campaign for reminders to DPG coordination and all DPGs will start  Find developer & make sure that the existing validator can take care of

upgrade validations

➢ Campaign on POG will follow, once the software is there.

https://twiki.cern.ch/twiki/bin/view/CMS/PdmVUpgradeValidation (developing...)

PPD Meeting, PdmV report, J-R Vlimant

Events Looking Different Feature

• Reported last week by Boris:

https://indico.cern.ch/conferenceDisplay.py?confId=169318

• Origin of the problem identified:

https://hypernews.cern.ch/HyperNews/CMS/get/recoDevelopment/1162/1.html

• Every single events (MC & Data) processed 5.2 and 5.3 are affected !

✗ Undetectable without looking at the event from two different datasets

• Fix for the bug available and will get into 6.X

✗ PFDisplacedVertexCandidateProducer use PVs from the previous event ✗ Get used in particle flow ✗ Get used in “rho” calculation

• Special build for 5.3 to address the amplitude of the issue

✔ Expected to be very small from developers ✔ Minimal samples should be available today ✔ Larger production will be done if effect is large.

PPD Meeting, PdmV report, J-R Vlimant

HCal Laser Pollution

https://twiki.cern.ch/twiki/bin/view/CMS/PdmVDataReprocessing534p2laserHCAL

• Solving with lumisection masking (certification) will amount to too much data thrown away
• Full list of events affected by the laser firing during collision needs to be done. Implies

running a skimming on every single /RAW Physics datasets of 2012

• Requests were send (Oct 24)

http://cms-pdmv-stats.web.cern.ch/cms-pdmv-stats/?search=franzoni_laserHCALskim

✔ Fast on processing ✗ Overhead in I/O ✔ 63/80 announced ✔ Expecting the remaining 13 (4 of which might not be relevant) soon

• >~600K events are affected
• Fraction of event polluted (see next slide)

➔ Average 0.02% pollution ➔ HCalNZS ~80% (no worries) ➔ Up to 1% or 2% in 2012A due to a bias of the first run, not black listed in the skim

• BTag, MET, Photon, SingleMu, MB,...
• Ready for making the full list, once we get the remaining datasets
• Not so sure about the format in which to deliver the list...
• Dataset available

➢ /RAW-RECO produced and can be looked at by Analysis groups ➢ AOD @ CERN

PPD Meeting, PdmV report, J-R Vlimant

HCal Laser Pollution Fraction

• Average 0.02% pollution
• HCalNZS ~80% (no worries)
• Up to 1% or 2%

➔ BTag, MET, Photon, SingleMu, MB,...

% 1% # in skim #in /RAW No corrected for DCS only efficiency