GlueX Offline Meeting, July 22, 2015
GlueX Offline Software Meeting
Wednesday, July 22, 2015
1:30 pm EDT
JLab: CEBAF Center F326/327
- Turning the Fine-Mesh Field Map into a Resource (Sean)
- Moving the Plug-Ins (Paul, Sean)
- Instructions for Setting Up the GlueX Environment (Mark)
- Nightly Builds Working from the Git Repository (Mark)
- SQL Injection Vulnerability, Offline Monitoring Pages (Justin)
- Review of minutes from July 8 (all)
- Offline Monitoring (Kei et al.)
- Spring 2015 Commissioning Simulations (Sean)
- Geant4 Update (Richard, David)
- Mixing Simulated Events (Sean)
- Conversion from Subversion to Git (Mark)
- ROOT 6 (Beni)
- 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/2015 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2015/ .
- BlueJeans: Kei Moriya
- CMU: Curtis Meyer
- FSU: Aristeidis Tsaris
- JLab: Alex Barnes, Amber Boehnlein, Mark Ito (chair), David Lawrence, Paul Mattione, Eric Pooser, Matt Shepherd, Nathan Sparks, Simon Taylor, Beni Zihlmann
- NU: Sean Dobbs
Find a recording of this meeting at https://bluejeans.com/s/8qJ6/ .
- Turning the Fine-Mesh Field Map into a Resource. Sean led us through his email. Before this change the fine-mesh map was computed on the fly if not present and then stored in the current working directory. This change allows the user to put the files in a single configurable location. If there is no fine-mesh field for a given field map file, the system falls back to the old procedure.
- Moving the Plug-Ins. We went over the emails from Paul and Sean. David had to do some work to modify the SCons scheme to build the plug-ins in their new location.
- Instructions for Setting Up the GlueX Environment. Mark reminded us about the wiki page he announced earlier. Paul has recently added a list of environment variables that the user might want to set that are not set by the general use scripts.
- Nightly Builds Working from the Git Repositories. The scripts have been modified to get the latest versions of HDDS and sim-recon using Git. They use a local clone of the GitHub repositories.
- SQL Injection Vulnerability, Offline Monitoring Pages. A team from DOE probing for cyber security vulnerabilities performed a successful SQL injection attack against a couple of the Offline Monitoring CGI Python scripts. Justin had the scripts fixed on the same day and they are back up and running now.
Kei gave a brief report:
- The last launch was two Fridays ago.
- The next launch is scheduled for this Friday.
- He plans to run a hybrid system of SWIF and jproj for this launch, taking advantage of SWIF's faster job submission mechanism. He is working with Chris Larrieu of SciComp on this.
- For the launch after that he hopes to have converted to an all-SWIF system.
- He plans to give an update on the conversion process and the final product at the next Offline Meeting. His goal is to produce something that others can use for their farm projects.
Sean reported that Marty Wise has fixed a problem on our database server (hallddb.jlab.org) that was causing the offline monitoring database access to hang. In the process Marty identified some queries that could use some optimization. Sean also mentioned that he and Justin are looking into whether the database tables need to be re-factored.
Spring 2015 Commissioning Simulations
Sean gave the report. Mark has done a first pass through running all of the jobs listed on the wiki page. He is filling in jobs that got stuck in bggen and timed-out without producing output.
Sean planning to run the same monitoring jobs on these files as is used in the offline monitoring jobs.
David gave the update. He has extracted some of the key classes from Richard's HDGeant4 repository and incorporated them in the CPP simulation he has been working on, in particular those that handle the geometry. Right now execution is very fast, but he is not sure that all of the material is seen by the generated particles. He suspects there is some subtlety having to do with Geant4's "parallel worlds" feature.
Beni asked if was possible to run HDGeant4 and pipe the output to mcsmear. David told us that he thought that problem has already been solved for HDGEANT, but he has not tried it himself.
Mixing Simulated Events
Sean described his work on mixing electromagnetic background (EMB) events with hadronic events from bggen. The hope is that time can be saved during simulation if the EMB does not have to be generated for every event.
He got David's hddm_merge_events program running. The program does not attempt to combine hits that are close in time when doing the mixing. He did comparisons between events that were generated with EMB in the standard way (where combination is done) and events where pre-generated EMB events were mixed with clean bggen events. In most cases the agreement is pretty good, within a few percent or better. The approach looks promising but there are still some unanswered questions. See his slides for details.
We discussed using real EMB events rather than simulated ones. This would likely give much more realistic simulated events.
- Currently, we could use pair spectrometer triggers that we already have.
- A dedicated random trigger would probably be better. There is a question of whether or not it needs to be synchronized with the machine RF.
- Right now there is a practical difficulty: real data is in EVIO format and the simulated events are in HDDM format. The mixing program works when both sources are HDDM. Sean and David will discuss a way around this.
Conversion from Subversion to Git
Mark summarized the conversion we did last week. In particular he went over new items on the wiki page documenting the process. Among the items he mentioned were:
- There is local copy of the Git repositories for hdds and sim-recon on our group disk at JLab. These are updated at noon and midnight and are now used to support the nightly builds.
- He posted some notes on the details of the conversion process to serve as a guide for future conversion efforts.
Sean asked about how we approach conversion of other directories from Subversion to Git in the future. At one point Mark proposed that we discuss any such conversion at this meeting and have the group decide on a case-by-case basis. We decided to adopt this as our policy.
David recently pushed changes directly to the master branch of the sim-recon repository at GitHub (i. e., no pull request). It appears that we do not get notifications when this happens like we do with pull requests. At least no one at the meeting got a notification. Mark will look into why this happened.
Beni has succeeded in building our software (under Ubuntu) using ROOT 6. He has already checked in the necessary changes to sim-recon. The system is still compatible with ROOT 5; the code senses which version is being used. He also had to make changes to JANA, but did not check those changes into the JANA repository.
GCC version 4.8 is required. The JLab batch farm is at 4.4 so we cannot use ROOT 6 for large-scale production at the Lab.
We thought we should go ahead and complete the work needed to have a ROOT-6-compatible system. We will continue to try and maintain compatibility with ROOT 5 so we don't lock ourselves out of the JLab Farm.
Action Item Review
- Work out details of doing an HDDM/EVIO merge.
- Complete work on changes to be compatible with ROOT 6.
- Figure out why we did not get notifications of a push directly to the master branch on GitHub.