(back to the list of all Panda meeting minutes)
Zoom Link
Time
8 am PT
Attendees
Fabio Hernandez Jen Adelman-Mccarthy Brian Yanny Edward Karavakis Michelle Gower Mikolaj Kowalik Peter Love Richard Dubois Wen Guan Zhaoyu Yang Wei Yang
Regrets
Agenda:
- Testing ci_hsc_gen3 at USDF and UKDF
- USDF: run all 126 pipeline tasks: https://panda-doma.cern.ch/tasks/?reqid=3163&display_limit=300&days=30
- UKDF: test with a single tasks (isr): https://panda-doma.cern.ch/tasks/?reqid=3316&days=30 (using gfal to workaround webdav issue, will test full/126 pipeline tasks)
- FrDF: waiting for a Butler setup
- what is the next step? trying to repeat DP0.2 at FrDF using Panda ? Or it is too ambitious (and need a smaller task to debug?)
- Docs Update
- Panda submission to DFs
- Panda service installation at USDF
- AOB
Notes:
- Full ci_hsc_gen3 testing jobs are also successful at UKDF https://panda-doma.cern.ch/tasks/?reqid=3360&display_limit=300
- Zhaoyu reported significant improvement of Panda JEDI and iDDS turn around time, after tuning by Wen (increase pulling frequency).
- Peter promised to allocate 100 dedicated cores for Rubin Panda testing
- Fabio thought that a full DP0.2 takes weeks or months, and is not a good next step. Suggested to use one or a few tracks in DP0.2 for the next step of Panda test at multi-DF
- Wen noticed that when submitting to USDF via ARC CE, Panda only get ~1000 jobs. Need to investigate whether this is a resource limitation issue (Wei is restricting Panda jobs at USDF (via ARC CE) to only run on the Rome nodes, not Milano nodes). or a ARC CE scaling issue (in this case, we may need to add more ARC CE).
- Eddy is working on production DB for Panda at USDF. The DB will be deployed with cloud native method (so its it HA).
- Where are we in multi-DF scaling test, need to update the status spread-sheet