(back to the list of all Panda meeting minutes)

Time

8 am PT

Attendees

Mikolaj Kowalik Fabio Hernandez Wen Guan Edward Karavakis Michelle Gower Brian Yanny Zhaoyu Yang Peter Love Wei Yang James Chiang 

Regrets

Richard Dubois Jen Adelman-Mccarthy 

Agenda:

  1. Update
    1. bps panda and clustering 
    2. other bps panda submission
    3. IDF to DF submission
    4. site issues
      1. do we still have Squid/NAT issue ?
      2. status of the second ARC CE at USDF
    5. panda installation at USDF
  2. Next steps
    1. status of replication and butler ingestion at DFs for the 2.2i/defaults/test-med-1 collections in /sdf/data/rubin/repo/dc2 repo. 

Notes:

  1. Updates
    1. Michelle will pump up the priority of getting bps panda plugin to work with clustering. Was fixing other bugs in control_bps
    2. Zhaoyu: remote build (to create execution butler) to use "bps submit" to instead of "prun" when submitting jobs. Need to change a few lines of code. Michelle will review
    3. when Michelle is busy and there are ticket about bps panda plugin, who will take them? Wen: the Panda team will work on them
    4. HTTP proxy at USDF works now. Some strange connections issue to Panda server: Zhaoyu's observation: for very short jobs, pilot update the status to 'running', and then to 'finished', and then ask the panda server for the next job. But panda server got the status order wrong, and send a 'kill' to pilot. pilot kills the new job. Working with pilot developer.
      1. Related issue, USDF http proxy only allows to port 80 and 443 of the Panda server, but the panda server listens on port 25443. Does this mean none of the communication with panda server actually use http_proxy? and used NAT instead?
    5. Panda sever at USDF: release a new DB scheme for postgres, to be used at USDF
    6. VOMS sever issue: after the power outage, VOMS is down and data transfer between UKDF and FrDF (using RAL FTS) can not proceed. Given that a quick switch to use IAM for VOMS is unlikely, Fabio is looking to see if FrDF can host an instance of VOMS (VOMS DB is very small and relatively static but do need to sync if there are more than one instance of VOMS DB).
    7. IDF to USDF: uses 'bps submit workflow.yaml' instead of prun to submit jobs to Panda (tested with isr), for all DFs (DM-37850)DDMM-3785DM-3785
  2. Next steps:
    1. UKDF is copying DC2 data from FrDF using FTS server at RAL. 
    2. Feels that it would be better that all 3 DFs could create the same small butler for data in 2.2i/defaults/test-med-1. 
    3. FrDF has instructure, At USDF, Wei tried and successfully created such a small butler. Will need to repeat this at UKDF once the raw data is ready 
    4. Jim Chiang believes that once the butler is ready, the campaign management team know how to generate workflow.yaml and submit to Panda.