GlueX Offline Meeting, July 13, 2018
GlueX Offline Software Meeting
Friday, July 13, 2018
10:00 am EDT
JLab: CEBAF Center A110
BlueJeans: 968 592 007
- 1 Agenda
- 2 Communication Information
- 3 Minutes
- Review of minutes from the June 29 meeting (all)
- Report from the HDGeant4 Meeting (Mark)
- Splitting up Sim-Recon (Sean, Mark)
- MC Request Form (Thomas)
- New osg production status monitor (Richard)
- Review of recent pull requests (all)
- Review of recent discussion on the GlueX Software Help List (all)
- Action Item Review (all)
- 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/2018 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2018/ .
- CMU: Naomi Jarvis
- JLab: Thomas Britton, Mark Ito (chair), Simon Taylor, Beni Zihlmann
- UConn: Richard Jones
- Yerevan: Hrach Marukyan
There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.
- New sim-recon version: 2.27.0, new HDGeant4 version: 1.7.0, new version set: 2.37. These updates were meant to provide a basis for HDGeant3 vs. HDGeant4 comparisons.
- Version for 2017 simulations: sim-recon recon-2017_01-ver02-sim_ver04. This version set is for simulation of Spring 2017 data and restores generation of CDC hits missing in the previous version. Simulation of bggen events using this version set will begin next week.
- We agreed that, in the future, we should have a definite commitment by an individual to check that a large simulation launch is producing reasonable results.
Review of minutes from the June 29 meeting
We went over the minutes from the last meeting.
The reproducibility problem reported last time has been solved. Simon reported that it was caused by variation in the order of hits presented to tracking. It was solved by using "stable sort" to guarantee the same ordering of hits on each run of the reconstruction.
Beni reported another problem where (not reproducibly) a block of 40 events can get lost upon input from an EVIO event source. This will get fixed in the re-write of the JANA package.
Report from the HDGeant4 Meeting
Mark reviewed his notes from the meeting on July 6. Please see the notes for details.
Since the meeting Thomas and Justin got together and decided on a set of nine single particle gun configurations (protons, π−s, and photons at various angle/energy combinations) along with the standard signal Monte Carlo samples (ρ, φ, ω, π0). A start at generating these samples has been made.
Splitting up Sim-Recon
Last weekend, Mark announced a test build of the split system. He received feedback from one taker, Justin Stevens, who reported success with the build.
Richard asked about modification of the HDGeant4 build. Mark has already taken care of that, on a branch in the repository. The same story applies to gluex_root_analysis.
To transition from sim-recon (unified system) to halld_sim and halld_recon (split system), we will have to change several things at the same time (or nearly the same time). We agreed on the following schedule.
- Early Monday, July 16: deploy the new version of build_scripts as the default at JLab. This version supports both the unified and split systems and should require no changes by users.
- Close of business, Tuesday, July 17: deadline for pushes to the unified system. Changes not committed and pushed to the unified system will not make it to the split system. Also note that for the halld_sim directories, they must appear on the master branch as well. All existing branches in sim-recon will appear in halld_recon.
- Early morning Wednesday, July 18: default configuration at JLab changes to split system.
MC Request Form
The system is backed up with a database. Requests are recorded and records of each individual job submitted are also kept. This will enable a richer variety of reports to users such as job status, resource usage, etc.
New OSG production status monitor (aka more computing resources for GlueX)
Richard has rolled out a new web page that monitors global job usage on the various university clusters available for running our OSG jobs. Since the collaboration meeting he has captured eight such clusters and has observed peaks of over 4000 simultaneous jobs, excluding those running on the OSG proper. The system will automatically switch jobs from site to site if any particular site gets backed up or goes offline. We first discussed Richard's proposal for adding university clusters at the June 1 Offline Meeting.
This is a significant new resource that can be accessed by GlueX OSG users with no special action other than specifying "wantjobrouter=true" in the job configuration. Other clusters may be added in the future.