BiLD-Dev

Europe/Zurich
2-R-014 (CERN)

2-R-014

CERN

Federico Stagni (Conseil Europeen Recherche Nucl. (CERN)-Unknown-Unknown)
Description
Bi-Weekly "Loyal" DIRAC developers meeting. And, following, the LHCbDIRAC developers meeting.

At CERN: Federico, Zoltan, Christophe, Andre', Vladimir, Alexander 
On Vidyo: Ueda, Hideki, Wojciech, Daniela, Simon, Andrei, Marcelo

Apologies: 

Follow-up from previous meeting

  •  

DIRAC communities roundtable

  • CLIC:
    • Mostly getting rid of deprecated functions
  • LHCb: 
    • Working on a LHCbDIRAC release still based on v6r20
      • Parametric jobs
      • Better platform selection
      • New ES DB (LHCb only)
    • Logs aggregation via MQ and LogStash
      • stopped momentarily, internal infrastructure issue, no DIRAC problem
    • Trying to understand why CS instances get stuck, need to be restarted regularly
      • File Descriptor issues
      • Most probably NOT related to DIRAC but rather to the infrastructure
  • GridPP:
    • v6r20p20 on test server, up to now looks fine, running v6r20p14
    • Migrating T2K from LFC to DFC
      • several issues to be be fixed, unrelated to DIRAC
      • Using tape at RAL (Castor)
      • Not yet managed to use, probably just need to install Storage Management System
  • Belle2:
    • Still running very old v6r17
      • going to upgrade ~soon
    • Asking if there have been changes in calculating priority for TQs
      • probably not, unless using ElasticSearch
      • Anyway several changes in TQs and SiteDirectors
      • Suggestion to use different groups for boosting the priorities of specific productions
  • EGI
    • Struggling with incorporation of SSO
    • WebApp DIRAC updated to include SSO login
      • Single sign-on check in possible
    • Still several open questions

Current situation

DIRAC

  • v6r20p22 latest production release
    • Mostly some fixes in SiteDirector
  • v6r20p21 buggy, will be retired

WebApp:

  • PRs:
    • 307: fixes problem introduced with SSO PR, so that host certificate were not anymore allowed 
  • v4r0 should be created now

Pilot3:

  • Wojtek restarted work on pilots logging together with a new student
    • Pilot itself is fine, the server is being prototyped right now as a tornado web app
    • More news soon

DIRACOS:

  • Shrinked size: now down to 250MB of bundle
    • removed some useless libraries (Java...)
    • devel removed
    • CVMFS deployment is anyway the way to go
  • DIRACOS v1r0 should be ready, tag soon
  • Deployment
    • start by putting it on EOS Web server at CERN, "like" lcgBundle
    • tarball: pipeline can be used for the upload
    • then we'll need to think about CVMFS, but later

Documentation

  • website diracgrid.org
    • No progress in the last 2 weeks
  • PR for cfg file in readthedocs:
    • the dirac.cfg should is created here https://dirac.readthedocs.io/en/integration/AdministratorGuide/Configuration/ExampleConfig.html and contains a full example

BOINC:

  • NTR

DB12:

  • NTR

tornado and other externals

  • Simon made a PR on Chris repository concerning M2Crypto, still to be evaluated

Release planning, tests and certification

Release planning:

  • v6r20 series
    • NTR
  • v7r0 series
    • Last PRs in the pipeline (dirac-install++ and ProductionSystem) 
      • Can all be merged
    • Should install on DIRAC certification with:
      • DIRAC v7r0-preX
      • DIRACOS  v1r0
      • WebApp 4.0
      • releases.cfg should be modified accordingly
    • bashrc will need to be re-done, so a fresh installation is needed
      • Instructions should go to https://github.com/DIRACGrid/DIRAC/wiki/DIRAC-v7r0
    • All Jenkins tests should be updated

Certification process

  • NTR

Weekly development(s) focus

  • Hosting for DIRAC tarballs will move to EOS-based one
    • trying to do everything behind the scene, affecting no-one
    • Won't move too old bundles, will keep v6r17 for Belle2
  • MQs and MQ topics (see presentation)
    • Wojciech presented what's available
    • just one question on structure and access (Queue vs Queues)
    • promised to improve documentation to explain how to set them up
    • Andrei raised some concerns on CS structure, as both "Queues" and "Topics"  are present, and this does not obviously fit the RSS structure

DIRAC: current PRs and tasks being worked on, or topics from Google forum

NTR

AOB

  • Next BiLD-Dev meeting will be held in 2 weeks (13 December, last meeting of the year)

LHCbDIRAC

  • v9r3 plan in the wiki
    • some things maybe will not make it for this release -- wiki updated
  • dirac-install deployment on on CVMFS together with the lcgBundle is really the last thing missing to be fully certified
    • we need Joel for that
    • we need instructions to be added so that everyone can release
    • all tasks in https://its.cern.ch/jira/browse/LBCORE-1410
  • Plan to release on December 3
    • it all depends from what we get today
    • if we can't make it, we will maybe release in January
  • v9r4 plan added to the wiki
    • One more release based on v6r20 with many LHCbDIRAC things
There are minutes attached to this event. Show them.