120 likes | 244 Views
Offline Production Readiness. Y.Fisyak STAR collaboration meeting. Outline:. Stability Chain status To do list. Stability. afs cache corruption on Linux server site increase memory on server (immediately) separate db and afs servers (September) client site increase afs cache to 1GB
E N D
Offline Production Readiness Y.Fisyak STAR collaboration meeting
Outline: • Stability • Chain status • To do list Y.Fisyak, STAR collaboration meeting
Stability • afs cache corruption on Linux • server site • increase memory on server (immediately) • separate db and afs servers (September) • client site • increase afs cache to 1GB • create afs cache on separate disk volume • set available cache 85% of volume size • compilers versions (egcs-1.1.2) • keep /usr/local/bin on STAR and RCF synchronized • populate /usr/local/bin both on RCF and STAR clusters via /afs/rhic/asis under STAR control Y.Fisyak, STAR collaboration meeting
Stability (cont.) • ROOT versions • new version coming ~ each month • new version usually associated with next .dev (dot - dev) STAR library release • there is backward compatibility but not always there is forward compatibility: • old root files can be read by new version but • not always new files can be read by old version • STAR Library versions usage • total no. of requests for SL ~297k • 98j - 32257; 98l - 7122: MDC1 • 99b (old) - 93123; MDC2 • 99d -18639; 99e (pro) - 33768; • 99f (new) -12134; 99g (dev) - 3655; 99h(.dev) - 314 Y.Fisyak, STAR collaboration meeting
Stability (cont.) • gmake to Perl (?) • a design goal to move in new environment two years ago (Jan. 1998) was to have • single • simple, and • easy to use Makefile. We have single and easy to use one but not simple. • The problem is that this Makefile is slow (too many dependencies) and it does not provide us with confidence of building complete set of libraries. • We study a possibility to use Perl replacement of gmake. Y.Fisyak, STAR collaboration meeting
Chain status • Macro bfc.C allows • to create chain with any kind of known Makers • define input/output • provide output in Root and xdf formats • self described • set of predefined chains for a given time stamp • tests: SD97, SD98 • simulation: year 1(a,b,c,d,e), year 2a, es99 • real data: SR99 • bfc.C ~1k locs but because it is stable enough it is a time to reduce it to ~100 locs by moving codes from bfc.C to compiled codes (StBFChain). Y.Fisyak, STAR collaboration meeting
Chain status (cont.) • DAQ ! • StDaqLib, thanks M.LeVine & co. and Iwona & co. • DAQ maker is a part StIOMaker (Victor). • St_tpcdaq_Maker reads DAQ and Trs output (H.Ward) • StEvent Display (Valery) • Systems: • TPC suffers from DB absence • SVT - srs, stk for 4th layer • CTB/TOF - codes should be fixed after engineering run (W.Llope). • EMC - a lot of developments (Akio, A.Pavlinov,...). Y.Matulenko the first time has demonstrated reconstructed π0 . Nothing new in the chain yet. Y.Fisyak, STAR collaboration meeting
Chain status (cont.) • Global • global maker has been turned to global chain (Wensheng Deng) whichincludes: • Match, • Primary, • V0, • Kink, and • Xi makers • Kalman fitter in global chain under testing(Al Saulys, Helen Caines) • if Kalman filter will adopted as baseline 3D fitter then GEANE will become essential part of reconstruction software. • QA exists and is working. A lot problems are found and fixed. (Kathy and 902 meeting). Y.Fisyak, STAR collaboration meeting
TO DO LIST • QA still on very qualitative level, to make itquantitative: • resolutions • efficiencies is necessary to include in QA StMcEvent. • DB’s: • production DB exists (Torre) • it is needed to build software to generate CRS jobs, • QA information should be part of this DB September (Lidia, W.Deng) • calibration/parameter DB (Jeff ?) • attempt to use egcs-1.1.2 on Sun fails due to inconsistency with CERN libraries. Solution is to make mini CERNLIB and GEANT3 using STAR environment. Y.Fisyak, STAR collaboration meeting
TO DO LIST(new dst tables and StEvent) • New dst tables (Lanny) • freeze new/pro and • make dev available for new tables • New StEvent: • reflects new dst tables • Structural and reference containers • organization hits/cluster in detector elements Data Sets • separate “Detector Reconstructed Tracks” (dst_track) -- seeds for Kalman -- and kinematics fits (multiple mass hypothesizes), V0, Xi, etc • relation among DST, mDST and μDST. Y.Fisyak, STAR collaboration meeting
StEvent • Goals are: • to store StEvent as DST • mDST is part of StEvent • to move filling of StEvent upstream in production chain • Wednesday afternoon time slot to discuss these topics Y.Fisyak, STAR collaboration meeting