Presentation is loading. Please wait.

Presentation is loading. Please wait.

Physics selection: online changes & QA M Floris, JF Grosse-Oetringhaus Weekly offline meeting 30/01/2011 1.

Similar presentations


Presentation on theme: "Physics selection: online changes & QA M Floris, JF Grosse-Oetringhaus Weekly offline meeting 30/01/2011 1."— Presentation transcript:

1 Physics selection: online changes & QA M Floris, JF Grosse-Oetringhaus Weekly offline meeting 30/01/2011 1

2 The problem PbPb 2011 run  frequent changes of trigger configuration (trigger class names, offline selection...) Did not propagate quickly enough from the trunk to the aliroot version used for the QA/AOD (i.e. the release) The main issue is related to trigger class names (no name  no data) The configuration is written in the OADB Possible solutions: Use a custom configuration in the QA macros (by hand) OADB on alien, path containing rev number, path given on JDL QA with analysis tag (one may need 1 tag/day + additional delays) 2 12/10/11Alice Physics Forum Preferred https://savannah.cern.ch/bugs/?89170

3 OADB on alien Path contains the rev number, it’s written in JDL: SplitArguments = {"/alice/…/#alienfilename# alien:////…/oadb_rev12345"}; OADB size: ~ 29 MB (and counting), 80% centrality calibration What it solves: New trigger class names, improved selection with already existing cuts Improved centrality calibration What it does not solve: An unforeseen selection cut is needed: this requires changing the classes (e.g. in 2010 we realized in the middle of the run that parasitic collisions could be rejected with ZDC) Backward compatibility: if a new selection/calibration is implemented, this can in principle affect earlier runs (reprocessing old runs with the new OADB may change them) If a class in OADB changes in the trunk, the new OADB may not be compatible with the rev (can be solved by policy) 3 12/10/11Alice Physics Forum

4 Status Local test successful (reading from Grid, as on a WN) Job sent on the grid failed (all subjobs either expired or went into EIB) Forensics (Latchezar): “jobs which ran killed for high memory consumption” Local test w/ valgrind: no issue spotted We believe this is not related to OADB Request: make a full test on a run which was already reconstructed in the standard production “The tests can be run after we finish with the Pb processing.” Comments? Suggestions? Conclusions? 4 12/10/11Alice Physics Forum


Download ppt "Physics selection: online changes & QA M Floris, JF Grosse-Oetringhaus Weekly offline meeting 30/01/2011 1."

Similar presentations


Ads by Google