GlueX Offline Meeting, July 12, 2017
GlueX Offline Software Meeting
Wednesday, July 12, 2017
11:00 am EDT
JLab: CEBAF Center F326/327
- Review of minutes from the last meeting (all)
- mcsmear development branch (Sean)
- Running mini launches on the grid? (Sean, Mark)
- HDvis update (Dmitry, Thomas)
- Review of recent pull requests (all)
- Review of recent discussion on the GlueX Software Help List (all)
- 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/2017 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2017/ .
- CMU: Naomi Jarvis
- FIU: Mahmoud Kamel
- JLab: Alexander Austregesilo, Alex Barnes, Thomas Britton, Mark Ito (chair), Dmitry Romanov, Justin Stevens, Simon Taylor
- NU: Sean Dobbs
- Yerevan: Hrach Marukyan
There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.
- JANA 0.7.9 released We looked over David's email announcing the new release.
- version_2.13 now available in gluex container We went over Richard Jones's announcement of a new container for running GlueX jobs on the grid.
Review of minutes from the last meeting
We went over the minutes from June 28.
- Progress on the Reconstruction Launch. Alex A. reported that the second part of REST production for the Spring 2017 data is about 60% done. He received a message from Sandy explaining that SciComp is working on a system for adjusting priority on the farm in a more automated way. There have also been problems that required reboot of the job scheduler.
- CCDB-related crashes. Dmitry report that he is working on a bug fix to recently reported crashes at the end of jobs that use the CCDB. This is due to a conflict in object ownership resulting in multiple calls to a destructor. The fix should be ready soon and a new tag will be applied.
- GlueX simulation CPU need projections. Mark mentioned that David Lawrence gave a presentation at the last Physics WG Meeting explaining the situation. The PWG will look into re-visiting our estimates going forward.
mcsmear development branch
Sean reviewed recent email describing a scheme for developing new mcsmear parameters and algorithms while keeping the reconstruction code base consistent with that used in the current REST production pass. This work will be done on a special Git branch, recon-2017_01-ver01-batch01-mcsmear. He also wrote a wiki page, describing the details of maintaining the branch.
Running mini launches on the grid?
Sean noted that with high demand on the JLab farm, we might want to run the planned series of mini-launches (simulation) on the Grid. Turnaround might be significantly faster and we would gain valuable experience. He started by modifying the configuration we used for sim 1.2.1 to set up jobs. Some issues:
- We want to use the mcsmear development branch he described earlier in the meeting. That is not in the current container and may have to be updated from time to time as we get going.
- We will have to periodically update the SQLite version of the CCDB that is downloaded to all of the OSG sites. The details of the mechanism has not been worked out.
- We have to decide on the framework for job submission. He has been working with the modified shell scripts used for sim 1.2.1. Richard has written an example Python script that configures jobs, submits them, and reports job status (see his email, linked in the announcements). Thomas has proposed using MCwrapper as the front-end.
Thomas and Dmitry demo'ed the latest version of the event display.
- The underlying rendering engine has been changed from ROOT's EVE to three.js. The event display runs entirely in the web browser now.
- Since the last meeting, a JANA-based server has been written to read and analyze events in the standard way and to send events to the browser in JSON format. This is a major milestone.
- The coding has been done to translate "raw" hits from all detector subsystems to JSON and thus these hits are currently available to the display program. In addition some higher level objects (e. g., tracks, FCAL clusters) have been translated. Not all of these have rendering code written for them (i. e., the program has the data, but does not know how to draw them). The plan is to do some sort of provisional rendering of all low-level objects and then turn things over to the individual detector groups to develop further rendering ideas (at least at the conceptual level), modifying the scheme for low-level objects and adding the important high-level ones.