meeting 2023 03 02

Notes for 2023-03-02 meeting

• date & time: Thu 2 Mar 2023 - 14:00 CEST (13:00 UTC)
  • (every first Thursday of the month)
• venue: (online, see mail for meeting link, or ask in Slack)
• agenda:
  • Quick introduction by new people
  • EESSI-related meetings in last month (incl. BioHackathon report)
  • First experiments with RISC-V
  • Progress update per EESSI layer (incl. bot for software layer)
  • 2021.12 version of pilot repository + outlook to next pilot version
  • AWS/Azure sponsorship update + OCRE funding opportunity
  • Update on MultiXscale EuroHPC project
  • Upcoming events
  • Q&A

Slides

• https://github.com/EESSI/meetings/blob/main/meetings/EESSI_meeting_20230302.pdf

Meeting notes

(by Kenneth)

Quick introduction by new people

• No new people

EESSI-related meetings in last month

(see slides)

• Collecting questions for the next CVMFS coordination meeting: https://hackmd.io/7otdq5tSRuyIWAJaF7NH6A
  • Usage tracking addressed, more later
  • Possible to Stratum 1 to create a Stratum 1, requires an opt-in
• Meetings with AWS/Azure/MultiXscale
  • AWS meeting next week important as credits will expire soon
  • Notes available for all (see slides)

Progress update per EESSI layer

Filesystem layer

(see slides)

• How will we do removal of versions in the long run?
  • There are a lot of options/ideas here (tarball, squashfs,...)
  • Nothing decided yet

Compatibility layer

(see slides)

• Some changes in bootstrap script that can cause the bootstrap to fail
• libc.so script now has file paths without prefix path, but requires that --sysroot is passed down to linker
• we should use older OS + kernel to bootstrap compat layer
  • but using older kernel also exposes more problems in bootstrap
• set up focused working group on compat layer?
• there's a GSoC'23 project to improve CI infrastructure for Gentoo Prefix: https://wiki.gentoo.org/wiki/Google_Summer_of_Code/2023/Ideas#Better_testing_infrastructure_for_Gentoo_Prefix

Software layer

(see slides)

• some issues opened by Alan for GPU-related stuff to fix/change in eessi_container.sh script
• CUDA is being "promoted" from runtime to build-only dependency (via EasyBuild hooks)
  • design choice, you won't see CUDA module being loaded even when software is GPU-aware
• GPU drivers are accessible in container via $LD_LIBRARY_PATH (in .singularity directory that is automatically created)
  • Apptainer bases this on output of ldconfig -p
  • non-priviledged Apptainer install can be included in compat layer (like Alliance is doing)

Usage "tracking" of EESSI

• Caspar: any privacy concerns w.r.t. collecting data like this?
  • could/should check with legal department of our institution?
• Terje: it would be (too) easy to feed fake data into counting server...
  • may need to not use UDP to partilaly avoid making this too easy?
• Alex: this may need to (always) to opt-in to be on the safe side w.r.t. data collection

Build-and-deploy bot

(see slides)

• Terje: largst missing feature right now is being able to request rebuild of a specific CPU target
  • would also be nice to "communicate" with bot through other means than labels (via comments)
  • bot should hide its own old comments, to help the human who can't otherwise see the forest through the trees (comments) anymore...

EESSI pilot repository

(see slides)

AWS/Azure sponsored credits

(see slides)

MultiXscale EU project

(see slides)

• open position at HPC-UGent: https://www.ugent.be/hpc/en/news-events/news/job-multixscale

Upcoming events

• HPCKP'23: https://hpckp.org/annual-meeting
  • submission deadline 7 April 2023
• BioHackathon Europe 2023
  • deadline for project proposals 10 April 2023

Q&A

• ...