Repository navigation
Make the TPC time-series task follow the active detectors - #2141
Conversation
|
REQUEST FOR PRODUCTION RELEASES: This will add The following labels are available |
|
@shahor02, @miranov25 : This should make the TPC time series task react on a disabled TOF detector. However, the workflow then actually crashes: It still searches for TOF. Could you maybe take a look when you have moment? |
sawenzel
changed the title
|
@matthias-kleiner, could you please review this Pull Request from @sawenzel? It appears to be a technical fix for the performance generator module and should not impact standard data taking or physics workflows. Thank you for checking! Best, |
| timeseries_clustertypes=cleanDetectorInputList("FT0,TOF,TPC") | ||
| TPCTStask = createTask(name='tpctimeseries_'+str(tf), needs=tpctsneeds, tf=tf, cwd=timeframeworkdir, lab=["RECO"], mem='2000', cpu='1') | ||
| TPCTStask['cmd'] = 'o2-global-track-cluster-reader --disable-mc --cluster-types "FT0,TOF,TPC" --track-types "ITS,TPC,ITS-TPC,ITS-TPC-TOF,ITS-TPC-TRD-TOF"' | ||
| TPCTStask['cmd'] = f'o2-global-track-cluster-reader --disable-mc --cluster-types {timeseries_clustertypes} --track-types {timeseries_tracktypes}' |
There was a problem hiding this comment.
It initialises unconditionally the readers for TOF clusters and tracks, in the absence of TOF this should lead to a crash. The timeseries_tracktypes and timeseries_clustertypes should not have contributions from detectors which are absent. Also the o2-tpc-time-series-workflow currently implicitly expects all track types it can digest, the same --track-types {timeseries_tracktypes} should be passed to it.
But why are the inputs of this workflow not simply copied from the anchor production workflow?
@matthias-kleiner, in the long run it would be better to modify the o2-global-track-cluster-reader to do what all other global workflows do: configure its input readers internally according to requested options, unless --disable-root-input is requested, e.g. https://github.com/AliceO2Group/AliceO2/blob/166180af16ecac7fa234714cdd7ae0394ea5a192/Detectors/GlobalTrackingWorkflow/src/SecondaryVertexingSpec.cxx#L246-L273
There was a problem hiding this comment.
Hello @shahor02 ,
many thanks for the detailed comment. I will take a look into it.
fe7021f to
6b34fe0
Compare
sawenzel
changed the title
This fixes the TRD/CTRACKLETS failure of the TPC time-series task in the MC workflow and makes its inputs follow the active detectors. - The reader used fixed cluster types without TRD, so the time series failed with "No matching output found for TRD/CTRACKLETS/0" once it requested TRD tracklets. - Track types now follow the active detectors and the anchored TRD and TOF matching sources. - The same track list goes to o2-tpc-time-series-workflow via --track-sources. - Builds with their own readers get --disable-root-input, so only one reader list exists. - TOF and TRD are optional, and the task requires ITS, TPC and FT0. - Builds without --track-sources keep the previous command and require TOF and TRD. https://its.cern.ch/jira/browse/O2-7264 AliceO2Group#2532 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
6b34fe0 to
1f8c321
Compare
This fixes the TRD/CTRACKLETS failure of the TPC time-series task in the MC workflow and makes its inputs follow the active detectors.
https://its.cern.ch/jira/browse/O2-7264
#2532
Tested with anchored MC on run 568012 (Pb-Pb 2025, apass2) with MC-prod-2026-v13-1 and v14-1, using async-2025-PbPb-apass2-v7 for reconstruction. Master fails with the TRD/CTRACKLETS error, while this version runs with all detectors, without TRD, without TOF and without both. On v14-1 alone, whose time series has its own readers, the output agrees with master within the sampling spread. For async-2024-pp-apass1-v7, which has no --track-sources, the command is identical to master and the task is skipped without TOF.
Assisted by Claude Code.