GlueX Software Meeting, November 13, 2018

From GlueXWiki
Jump to: navigation, search

GlueX Offline Software Meeting
Tuesday, November 13, 2018
3:00 pm EST
JLab: CEBAF Center A110
BlueJeans: 968 592 007

Agenda

  1. Announcements
    1. scosg16 upgraded.
    2. MCwrapper v2.0.3 released
      1. Reaction lines demo
  2. Review of minutes from the October 30 meeting (all)
  3. Review of recent issues and pull requests:
    1. halld_recon
    2. halld_sim
    3. hdgeant4
  4. Review of recent discussion on the GlueX Software Help List (all)
  5. Action Item Review (all)

Communication Information

Remote Connection

Slides

Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2018 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2018/ .

Minutes

Present:

  • CMU: Naomi Jarvis
  • FSU: Sean Dobbs
  • JLab: Alexander Austregesilo, Stuart Fegan, Colin Gleason, Mark Ito (chair), David Lawrence, Simon Taylor, Beni Zihlmann
  • Regina: Ahmed Foda
  • W&M: Justin Stevens

There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.

Announcements

  1. scosg16 upgraded. Thomas reported that the upgrade of the network card on scosg16.jlab.org solved the problem he has been fighting with slow job submission to the OSG and lost access to our database server. The card was upgraded from 1 Gbit/s to two bound 10 Gbit/s interfaces, a 20-fold increase in bandwidth. The problems, before changing the interface, was due to shipping random trigger files to the various OSG sites. Where it used to take 6 to 8 hours to submit 2,500 jobs, it now takes 30 minutes.
  2. MCwrapper v2.0.3 released. See the release notes for complete information. Thomas has implemented a reaction line composer to the MCwrapper submit page. DReactions that users specify will be produced as ROOT trees along with the output REST files.
  3. Offline Monitoring Launch Job Crashes. Alex A. reported that the last monitoring launch had only a 50% yield. He tried running the code used in last NERSC campaign with a nearly 100% success rate. There is a suspicion that a change introduced some weeks ago to remove spurious tracks might be the cause. Alex will experiment with a build that excludes that change to see if it makes a difference.
    • Simon noted that the root cause of the spurious tracks were candidates that changed sign after re-fits done during reconstruction. These in turn resulted in tracks that had two occurrences of a particular particle type (say π) which in turn resulted in apparent "duplicate" combos.

Review of minutes from the previous meeting

We went over the minutes from the October 30 meeting.

  • Computing Review. Curtis and David have posted draft slides to the Indico site.
  • NERSC. David has received the green light for another monitoring launch at NERSC, using version 19. He ran into a SWIF2 problem that Chris Larrieu has addressed. The idea is to analyze all runs, using 5% of the files for each.
  • Work Packages. No progress has been made assigning names to packages.

Review of recent issues and pull requests

halld_recon

  • Issue #40: Sean broke the tracking again... Problem seen with ρ yield. Sean traced it to accidental commenting out of a line of code. Not clear how it worked at all. Problem was really not with the ρ yield; rather the normalization against number of events.
  • Issue #39: Problems with kinematic fitter in Lambda events? There was a lot of discussion on this issue, but in summary:
    • Sean and Beni found NaNs in the fitting. These came from tracks coming from way upstream or from tracks with huge x and/or y starting coordinates. This problem did not dominate and are not related to the large deficits that Hao Li is reporting.
    • Alex's report of 50% of jobs crashing on an analysis launch with Hao's fix branch may be related to the crashes he reported earlier in the meeting when using the most recent code (see the announcements above). He will investigate.

halld_sim

  • Issue #16: mcsmear crash. This sparked a discussion about RCDB support and responsibility for (a) maintaining the code (b) correct input of critical values during running. For (a) Dmitry Romanov is here now and we are hoping he will be able to devote significant time to RCDB and for (b) we should contact the online group to make sure all needed parameters are correctly added. [added in press: see Alex's comment on this issue about conversations with Sergey Furletov and Dmitry.]

hdgeant4

There has been very little work in this area lately. Simon has started looking at the problem of FDC hits disappearing from tracks.