GlueX Offline Meeting, September 14, 2016
GlueX Offline Software Meeting
Wednesday, September 14, 2016
2:00 pm EDT
JLab: CEBAF Center F326/327
- Review of minutes from the last meeting (all)
- Launches (Paul/Alex A./Sean)
- Reconstruction Launch
- Software Review
- Review of recent pull requests (all)
- Review of recent discussion on the Gluex Software Help List.
- Action Item Review
- The BlueJeans meeting number is 968 592 007 .
- Join the Meeting via BlueJeans
Talks can be deposited in the directory
/group/halld/www/halldweb/html/talks/2016 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2016/ .
- CMU: Naomi Jarvis, Curtis Meyer, Mike Staib
- JLab: Alexander Austregesilo, Brad Cannon, Mark Ito (chair), David Lawrence, Paul Mattione, Nathan Sparks, Simon Taylor, Beni Zihlmann
- NU: Sean Dobbs
- Fix to JEventSource_EVIOpp crashes on PS skims. We looked over Richard Jones's bug fix to get around anomalies in the pair spectrometer skims. Sean has reviewed the problem and the change. With recent changes to the skimming software, it is possible that the anomalies have disappeared.
- sim-recon version 2.5.0. See the release notes in the link for details. At JLab this release has been build against ROOT 6.06.08. A build against ROOT 5 is coming soon.
Review of minutes from the last meeting
Alex reported progress on the reconstruction launch and its corresponding analysis launch.
The launch was scheduled to begin Friday, two weeks ago, but due to last-minute software changes and associated problems, it was delayed until Saturday. After Batch 1 had retrieved all of its files Alex started Batch 2, to keep the tape library busy.
Processing of these two batches took 9 days total (ending this past Monday). The memory footprint was a bit larger than in version 1. The requested amount had to be increased. Each job was run with 14 threads.
We looked at the Launch webpage which shows various statistics on the jobs. The success rate was 99.97%. The number of pending jobs never went much above 500. The number of running jobs sometimes dipped to zero.
The jobs for Batch 1 and 2 are running now. There are 50 analysis plugins, in addition to the standard ones, from 12 or 13 contributors. Of the 9 k, jobs, 7 k are finished.
We looked at this launch page as well. Because the input REST files were already resident on the cache disk, the number of active jobs has been stable at a few hundred.
When these jobs are finished, Alex will go back to reconstruction on Batches 3 and 4.
Curtis created a wiki page to collect information for review preparation. He also started writing a document that we will use to organize content for presentation at the review. He is soliciting input from the Collaboration on topics to be presented and, even better, paragraphs, tables, and figures to flesh out the sections.
Mark emphasized that the Launches should be featured prominently. Each launch has its own character and the statistics from each the launch tell a different story in each case. It is a chance to not just show projections, but to present performance of the software IRL.
We may try to put the document into a private repository in GitHub. Mark will look into this.
- Look into private repositories on GitHub. (Mark)
- Contribute paragraphs and figures to the Software Review document. (All)
- Read white paper on review of large farm projects. (All)