Difference between revisions of "GlueX Software Meeting, February 16, 2022"

From GlueXWiki
Jump to: navigation, search
(fill in agenda)
(minutes added)
 
Line 75: Line 75:
 
# Meeting Time in January (all)
 
# Meeting Time in January (all)
 
# Action Item Review (all)
 
# Action Item Review (all)
 +
 +
== Minutes ==
 +
 +
Present: Alex Austregesilo, Thomas Britton, Sean Dobbs, Sergey Furletov, Mark Ito (chair), Richard Jones, David Lawrence, Curtis Meyer, Churamani Paudel, Simon Taylor
 +
 +
There is a [https://jlab-org.zoomgov.com/rec/share/3F2LobRgURzg__3SZeObjqQstB0o4uGe9kwiaoG7pgZ_dnrK0abJrZ9wGSNoAnvf.i3Hz4BHsHS4FR6wR recording of this meeting] on the ZoomGov site. Passcode: ey1uBqA#
 +
 +
=== Announcements ===
 +
 +
# [https://mailman.jlab.org/pipermail/halld-offline/2022-February/008760.html New release of Build Scripts: version 2.31]
 +
#* Building ROOT from the master branch of the GitHub repository is supported, but Mark has not been able to get a successful build.
 +
# [https://mailman.jlab.org/pipermail/halld-offline/2022-February/008764.html New version set: 5.4.0] Update releases.
 +
# [https://mailman.jlab.org/pipermail/halld-offline/2022-February/008765.html New version set: recon-2019_11-ver01.2] Includes a bug fix to the creation of ROOT output trees.
 +
 +
=== Review of Minutes from the Last Software Meeting ===
 +
 +
We went over the [[GlueX Software Meeting, February 2, 2022#Minutes|minutes from the meeting on February 2]].
 +
 +
We had a discussion of several issues around transition to a new version of a Linux distribution.
 +
 +
* Docker has taken down its auto-build feature for non-paying accounts. We have been using this to produce containers for the OSG and the HPC centers. Richard reported that the "docker push" command on a Dockerfile can be used to initiate a build on DockerHub. We should move to that.
 +
** Mark remarked that this change in our procedures provides an opportunity to move our Docker containers from the "markito3" account to the "JeffersonLab" account on DockerHub.
 +
* The [https://hub.docker.com/_/centos/ CentOS account on DockerHub] does not provide a CentOS 8 stream container. However, it is being [https://hub.docker.com/r/tgagor/centos/tags provided by a private citizen]. Mark has not built a GlueX container from it yet.
 +
* Richard has a GlueX-ready CentOS 8 singularity container on CVMFS: /cvmfs/singularity.opensciencegrid.org/rjones30/gluexpro8:latest.
 +
* Mark is working on such a container that has all of the latest packages requires for the latest tags of our various packages.
 +
** The CentOS 8 Docker container has been deprecated such that it cannot be used to build a GlueX singularity container without hand modification.
 +
** Downloads of CentOS 8 RPM packages from the deprecated site are extremely slow.
 +
* Curtis is running CentOS 8 Stream on the cluster at CMU, an existence proof.
 +
* Richard is interested in a CentOS 8 container to debug issues reported by Naomi on CentOS 8 Stream.
 +
 +
Richard has addressed [https://github.com/JeffersonLab/HDGeant4/issues/204 hdgeant4 issue #204] by going to the CCDB to get info for beam parameters if a BEAM card in not present in the control.in. See [https://github.com/JeffersonLab/HDGeant4/pull/208 HDGeant4 pull request #208] and [https://github.com/JeffersonLab/halld_sim/pull/233 halld_sim pull request #233].
 +
 +
=== Report from the SciComp Meeting ===
 +
 +
Mark presented the [https://halldweb.jlab.org/talks/2022/sciops+enp-bhess-feb22.pdf slides that Bryan Hess showed] at the SciComp meeting on February 3rd. Present at that meeting besides Bryan and Mark were Amitoj Singh, Brad Sawatzky, David Lawrence, Markus Diefenthaler, Ole Hansen, Thomas Britton, and Nathan Baltzell. Please see Bryan's slides for details. Some related points:
 +
 +
* At the SciComp meeting, Mark did mentioned the idea of JLab providing StashCache space in addition to that already deployed from UConn as well as space-to-come from CMU. As noted in Bryan's slides, other than the disk space itself, JLab has all of the infrastructure needed to support StashCache up and running. At the meeting Bryan remarked that this is do-able without a lot of effort and promised to look into it.
 +
* Bryan has already agreed to come to our next meeting (March 2) to discuss status and plans for these OSG-related, outward facing (from a JLab perspective) systems.
 +
 +
=== FAQ of the Fortnight: What do I have be careful of when creating ROOT histograms and trees with JANA? ===
 +
 +
[[GlueX_Offline_FAQ#What_do_I_have_be_careful_of_when_creating_ROOT_histograms_and_trees_with_JANA.3F|This FAQ]] simply points to [[Locking in JANA|this wiki page]] which Sean had referenced during a recent [https://halldweb.jlab.org/wiki-private/index.php/January_25,_2022,_Open_Analysis_Discussion Open Analysis Discussion]. It is a golden oldie, from 2016, and provides guidance on how to handle ROOT histograms and trees in the multi-threaded environment of JANA.
 +
 +
There was some discussion:
 +
 +
* Related to the burden on users of managing locks for ROOT, David reported that JANA 2 is very close to a release. He will ping Nathan Brei on where we are.
 +
* David remarked that we probably should embark on a program to keep our software up-to-date with the latest versions of C++. In general we are C++11 compatible, but the world is already using C++20 as the standard (not to mention C++14 and C++17).
 +
* Mark remarked that although different from the C++ situation, we do have a situation with Python 2 vs. Python 3. This issue has come up over and over during the past few years and solutions have thus far been of the Band-Aid (TM) type. Sean mentioned the current situation with RCDB 0.07.00 as an example, one that needs to be addressed sooner rather than later.
 +
 +
=== Review of recent issues and pull requests ===
 +
 +
Sean called our attention to [halld_recon pull request #627] '''Explicitly check to see if we're at the end of the file, and return the proper status'''. It gets around a problem where the EVIO footer does not get written out. The root cause needs to be investigated, but this eliminates the hangs.

Latest revision as of 21:09, 16 February 2022

GlueX Software Meeting
Wednesday, February 16, 2022
1:30 pm EST

Zoom Meeting ID: 161 869 2159 Passcode: 634605 Join

Mark Ito is inviting you to a scheduled ZoomGov meeting.

Topic: GlueX Software Time: This is a recurring meeting Meet anytime

Join ZoomGov Meeting https://jlab-org.zoomgov.com/j/1618692159?pwd=VGpjT1BZR2hKWmp1S0E3aEpHUlRiQT09

Meeting ID: 161 869 2159 Passcode: 634605 One tap mobile +16692545252,,1618692159# US (San Jose) +16468287666,,1618692159# US (New York)

Dial by your location

       +1 669 254 5252 US (San Jose)
       +1 646 828 7666 US (New York)
       +1 669 216 1590 US (San Jose)
       +1 551 285 1373 US
       833 568 8864 US Toll-free

Meeting ID: 161 869 2159 Find your local number: https://jlab-org.zoomgov.com/u/acAwo1X4w9

Join by SIP 1618692159@sip.zoomgov.com

Join by H.323 161.199.138.10 (US West) 161.199.136.10 (US East) Meeting ID: 161 869 2159 Passcode: 634605

Agenda

  1. Announcements
    1. New release of Build Scripts: version 2.31
    2. New version set: 5.4.0
    3. New version set: recon-2019_11-ver01.2
  2. Review of Minutes from the Last Software Meeting (all)
  3. Report from the SciComp Meeting (Mark)
  4. FAQ of the Fortnight: What do I have be careful of when creating ROOT histograms and trees with JANA?
  5. Review of recent issues and pull requests:
    1. halld_recon
    2. halld_sim
    3. CCDB
    4. RCDB
    5. MCwrapper
    6. gluex_root_analysis
  6. Review of recent discussion on the GlueX Software Help List (all)
  7. Meeting Time in January (all)
  8. Action Item Review (all)

Minutes

Present: Alex Austregesilo, Thomas Britton, Sean Dobbs, Sergey Furletov, Mark Ito (chair), Richard Jones, David Lawrence, Curtis Meyer, Churamani Paudel, Simon Taylor

There is a recording of this meeting on the ZoomGov site. Passcode: ey1uBqA#

Announcements

  1. New release of Build Scripts: version 2.31
    • Building ROOT from the master branch of the GitHub repository is supported, but Mark has not been able to get a successful build.
  2. New version set: 5.4.0 Update releases.
  3. New version set: recon-2019_11-ver01.2 Includes a bug fix to the creation of ROOT output trees.

Review of Minutes from the Last Software Meeting

We went over the minutes from the meeting on February 2.

We had a discussion of several issues around transition to a new version of a Linux distribution.

  • Docker has taken down its auto-build feature for non-paying accounts. We have been using this to produce containers for the OSG and the HPC centers. Richard reported that the "docker push" command on a Dockerfile can be used to initiate a build on DockerHub. We should move to that.
    • Mark remarked that this change in our procedures provides an opportunity to move our Docker containers from the "markito3" account to the "JeffersonLab" account on DockerHub.
  • The CentOS account on DockerHub does not provide a CentOS 8 stream container. However, it is being provided by a private citizen. Mark has not built a GlueX container from it yet.
  • Richard has a GlueX-ready CentOS 8 singularity container on CVMFS: /cvmfs/singularity.opensciencegrid.org/rjones30/gluexpro8:latest.
  • Mark is working on such a container that has all of the latest packages requires for the latest tags of our various packages.
    • The CentOS 8 Docker container has been deprecated such that it cannot be used to build a GlueX singularity container without hand modification.
    • Downloads of CentOS 8 RPM packages from the deprecated site are extremely slow.
  • Curtis is running CentOS 8 Stream on the cluster at CMU, an existence proof.
  • Richard is interested in a CentOS 8 container to debug issues reported by Naomi on CentOS 8 Stream.

Richard has addressed hdgeant4 issue #204 by going to the CCDB to get info for beam parameters if a BEAM card in not present in the control.in. See HDGeant4 pull request #208 and halld_sim pull request #233.

Report from the SciComp Meeting

Mark presented the slides that Bryan Hess showed at the SciComp meeting on February 3rd. Present at that meeting besides Bryan and Mark were Amitoj Singh, Brad Sawatzky, David Lawrence, Markus Diefenthaler, Ole Hansen, Thomas Britton, and Nathan Baltzell. Please see Bryan's slides for details. Some related points:

  • At the SciComp meeting, Mark did mentioned the idea of JLab providing StashCache space in addition to that already deployed from UConn as well as space-to-come from CMU. As noted in Bryan's slides, other than the disk space itself, JLab has all of the infrastructure needed to support StashCache up and running. At the meeting Bryan remarked that this is do-able without a lot of effort and promised to look into it.
  • Bryan has already agreed to come to our next meeting (March 2) to discuss status and plans for these OSG-related, outward facing (from a JLab perspective) systems.

FAQ of the Fortnight: What do I have be careful of when creating ROOT histograms and trees with JANA?

This FAQ simply points to this wiki page which Sean had referenced during a recent Open Analysis Discussion. It is a golden oldie, from 2016, and provides guidance on how to handle ROOT histograms and trees in the multi-threaded environment of JANA.

There was some discussion:

  • Related to the burden on users of managing locks for ROOT, David reported that JANA 2 is very close to a release. He will ping Nathan Brei on where we are.
  • David remarked that we probably should embark on a program to keep our software up-to-date with the latest versions of C++. In general we are C++11 compatible, but the world is already using C++20 as the standard (not to mention C++14 and C++17).
  • Mark remarked that although different from the C++ situation, we do have a situation with Python 2 vs. Python 3. This issue has come up over and over during the past few years and solutions have thus far been of the Band-Aid (TM) type. Sean mentioned the current situation with RCDB 0.07.00 as an example, one that needs to be addressed sooner rather than later.

Review of recent issues and pull requests

Sean called our attention to [halld_recon pull request #627] Explicitly check to see if we're at the end of the file, and return the proper status. It gets around a problem where the EVIO footer does not get written out. The root cause needs to be investigated, but this eliminates the hangs.