[Halld-offline] Software Meeting Minutes, July 20, 2021

[Mark Ito]

People,

Please find the minutes here 
<https://halldweb.jlab.org/wiki/index.php/GlueX_Software_Meeting,_July_20,_2021#Minutes> 
and below.

   -- Mark

      _____________________________________


    GlueX Software Meeting, July 20, 2021, Minutes

Present: Alex Austregesilo, Sean Dobbs, Mark Ito (chair), Richard Jones, 
David Lawrence, Churamani Paudel, Simon Taylor, Jon Zarling, Beni Zihlmann

There is a recording of this meeting 
<https://bluejeans.com/s/Ksrj6a84hWJ>. Log into the BlueJeans site 
<https://jlab.bluejeans.com> first to gain access (use your JLab 
credentials).


      Announcements

 1. New work disk is up and running as a disk server. Files are being
    migrated from old server to new. Initially we will have about 400 TB
    on the new server compared to the 200 TB we have on the current
    /work/halld and 50 TB we have on /work/halld3. After the transition
    is complete, the old servers will have their operating systems
    upgraded and will be redeployed, so we will recover the old disk
    space in addition to the new.
 2. AmpTools v0.12.1
    <https://mailman.jlab.org/pipermail/halld-offline/2021-July/008582.html>.
    We reviewed Matt Shepherd's announcement. There is a new version set
    available that has this bug-fix release of AmpTools (version set 4.42.1)


      Review of Minutes from the Last Software Meeting

We went over the minutes from the meeting on July 6th 
<https://halldweb.jlab.org/wiki/index.php/GlueX_Software_Meeting,_July_6,_2021#Minutes>. 
Turns out we did have a work disk squeeze since the last meeting. Mark 
Dalton and George Vasileiadis helped out to reduce usage on /work/halld.


      Review of Minutes from the Last HDGeant4 Meeting

We reviewed the minutes from the meeting on July 13th 
<https://halldweb.jlab.org/wiki/index.php/HDGeant4_Meeting,_July_13,_2021#Minutes>. 
Richard has made progress on adding realism to the FDC efficiency 
modeling. See this Issue 
<https://github.com/JeffersonLab/HDGeant4/issues/181#issuecomment-884957518> 
on the HDGeant4 GitHub site.


      FAQ of the Fortnight

Mark introduced a new segment of the Software Meetings: FAQ of the 
Fortnight. At future meetings, he will lead us through one of the FAQ's 
(or similar pieces of documentation) in an effort to boost collaboration 
productivity. This week we reviewed Where do I find version set files? 
<https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_FAQ#Where_do_I_find_version_set_files.3F>. 
You can find the tour starting at 14:30 in the recording.


      DSelector Choices and Disk Footprints

Jon has been looking several ROOT-related and ROOT-adjacent software 
ideas with an eye toward reducing storage size of ROOT trees and 
improving I/O performance when reading them back. See his slides 
<https://halldweb.jlab.org/wiki/images/d/df/OfflineJZ_footprint_and_uproot_7.20.21.pdf> 
for all of the details. His conclusion slide reads:

  * Either ROOT::Math::LorentzVector or four doubles could reduce our
    footprint by about 50%
      o ROOT::Math::LorentzVector could probably be wrapped into
        DLorentzVector (replacing TLorentzVectors currently implemented)
  * Fiducial kinfit cuts could reduce the footprint even more (but might
    need analyzer input)
  * Using LZ4 compression (ROOT 6.24.00+) increases size by 15-20%
      o But can parse 1.5-2× faster
  * Working on a wrapper for uproot in flat tree format, stay tuned!


      System for using GCC 8 with the GlueX software stack

Mark reviewed his email message 
<https://mailman.jlab.org/pipermail/halld-offline/2021-July/008587.html> 
outlining a new system, ready for testing, for accessing a more advanced 
version of the GCC compiler that is shipped with CentOS 7. This allows 
access to more modern versions of our support software, in particular at 
present the latest version of Geant4.

There is a new HOWTO 
<https://halldweb.jlab.org/wiki/index.php/HOWTO_Run_Gluex_Software_with_GCC_8> 
on...well...how to use the system in a Singularity container. This 
method is currently available on the OSG and the HPC centers. Mark is 
working on other approaches including running CentOS 8 and using GCC 
"modules" at JLab.


      Review of recent issues and pull requests

We discussed halld_recon issue #537, Problems with photon energies in MC 
samples <https://github.com/JeffersonLab/halld_recon/issues/537> posted 
by Sean. Please see the issue for all of the details. This one is tough 
because it is hard to reproduce.

On another front, Sean also mentioned that Igal is having a 
non-reproducible issue with the last of the hundreds of jobs in a run 
hanging at the end of its file.


      Action Item Review

 1. Decide on needed tape volume sets. (Sean, Mark I.)
 2. Review disk usage when re-repatriating recon launch data. (Alex, Mark)
 3. Discuss automatically making CCDB entries for selected RCDB items.
    (Mark I., Sean)
 4. Modify the documentation on how to copy files from JLab. (Mark I.)
 5. Generate finer-grained info on small cache files. (Mark I.)


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20210726/90708892/attachment.html>