GlueX Containers Meeting, March 16, 2018

From GlueXWiki
Jump to: navigation, search

GlueX Containers Meeting
Friday, March 16, 2018
11:30 am EST
JLab: CEBAF Center F226
BlueJeans: 968 592 007



Present: Thomas Britton, Mark Ito, Sean Dobbs, David Lawrence, Richard Jones

  • Sean mentioned analysis launches as a candidate activity for grid/hpc
  • David: there is a big push to use NERSC for data reconstruction, issue is input (raw) data transport
  • Richard pointed out that when Edgar Farjado put together the Submit Host, we told him we were going to do mainly Monte Carlo. There might be system configuration parameters that should be changed (hardware or software) to support reconstruction of raw data. We might want to meet with Edgar again.
  • David got CVMFS going on his MacBook. "Very easy". Could not see the CVMFS partition from inside a Docker container. A worry for NERSC.
  • Mark and Thomas have still have to work on a system for transferring grid output files from the submit host to cache.
  • Thomas: random trigger files are rsynched to the /osgpool partition on the submit host and shipped with jobs via the Condor mechanism. Files are hundreds of megabytes.
  • Sean has mentioned Stash as an OSG mechanism for distributing input files. Random triggers are a good candidate for this.
  • Thomas: reconstruction could be done in MCwrapper
  • David mentioned a Docker swarm as a technique for managing our various containers