GlueX Software Meeting, July 23, 2019
GlueX Software Meeting
Tuesday, July 23, 2019
3:00 pm EDT
JLab: CEBAF Center A110
BlueJeans: 968 592 007
- 1 Agenda
- 2 Minutes
- 2.1 Announcements
- 2.2 Review of minutes from the last Software Meeting
- 2.3 Dropping RedHat Enterprise Linux 6 from list of supported distributions
- 2.4 New update release? AmpTools? RCDB?
- 2.5 Proposal for analysis launch and skimming
- 2.6 Fall 2018 Reconstruction Launch
- 2.7 Review of recent issues and pull requests
- Review of minutes from the last Software Meeting (all)
- Dropping RedHat Enterprise Linux 6 from list of supported distributions (Mark)
- New update release? AmpTools? RCDB?
- Proposal for analysis launch and skimming
- Review of recent issues and pull requests:
- Review of recent discussion on the GlueX Software Help List (all)
- Action Item Review (all)
Talks can be deposited in the directory
/group/halld/www/halldweb/html/talks/2019 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2019/ .
- FSU: Sean Dobbs
- JLab: Alexander Austregesilo, Mark Ito (chair), Simon Taylor
- Undisclosed Location: David Lawrence
There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.
- Cache disk outage on Aug. 8. The outage is to bring up a new Lustre system.
- We discussed the problem with the current write-through cache system with many small files, none of them backed up to tape.
- Launch News. There is a new Monitoring Launch 2019-01 ver05 in preparation
Review of minutes from the last Software Meeting
We went over the minutes from July 9. Preparations for the GlueX-PANDA ML workshop are continuing. David is in contact with Tobias Stockmanns on the agenda.
Dropping RedHat Enterprise Linux 6 from list of supported distributions
After the last maintenance period at JLab, jlabl3.jlab.org transitioned to RHEL7 and we lost our last publicly available RHEL6 system. Mark proposed that we stop building on that distribution, effectively stopping support. He listed use-cases that might suffer:
- Users of the few nodes on the gluon cluster that run RHEL6
- Desktop users at the lab that have not transitioned to RHEL7. Simon is not among them.
- Users outside the lab that depend on builds as a proof of continued viability on the distribution.
Mark pooled those present at the meeting and no one objected to dropping support. Mark will therefore discontinue RHEL6 builds. Nightly builds on RHEL6 have already stopped.
We also thought that support of CentOS6 could also be dropped.
New update release? AmpTools? RCDB?
Alex explained that the latest pre-release of AmpTools has a few new, useful features, including reduced memory usage and faster execution time. Use of GPUs has also improved. He has tested it and has not seen any problems. He encouraged its inclusion in the next update version set.
He also mentioned that we need a new version of RCDB so that a few new aliases, needed for processing DIRC and PrimEx data, can be used. This should also be included in a new version set.
Mark agreed to roll out a new version set soon.
Proposal for analysis launch and skimming
We briefly discussed Matt Shepherd's recent proposal to produce "flattened" analysis trees as part Analysis Launches. There was some substantial discussion on the offline list (threaded to the previous link). It seems that there is agreement that smaller data sets would be a good thing, but there is the possibility that modest tightening of cuts at analysis time itself might yield a large fraction of the possible gains. In addition if fewer "extra" particles are considered, kinematic fitting will go a lot faster (fewer combos to fit).
There will be more discussion of the idea at the Production and Analysis Meeting.
Fall 2018 Reconstruction Launch
Alex reported that he has tagged a version of the software for use in processing Fall 2018 data that differs from that used for Spring 2018 only by an update to the TPOL plugin, so essentially the same software.
David reported an outage at NERSC from July 26 through 30. The plan is to wait until after the outage before starting the launch. In the mean time, test jobs will be submitted to make sure we are ready.
Review of recent issues and pull requests
Sean reported that there are some fundamental structural issues with the analysis libraries that contribute to the crashes when analyzing multi-run files (halld_recon Issue #111). Mark Dalton ran into this problem recently analyzing J/ψ data. The recent fixes to how some plugins are initialized addresses some problems, but not all of them. This one will take some effort to disentangle.