Skip to content
Zack Galbreath edited this page May 12, 2023 · 1 revision

Attendees

  • Aashish Chaudhary
  • Alec Scott
  • Dan LaManna
  • Jacob Nesbitt
  • Massimiliano Culpo
  • Mike VanDenburgh
  • Todd Gamblin
  • Zack Galbreath

Rolling releases

  • We created and deployed a new cron job in our cluster. It is responsible for creating a tagged release on GitHub (name formatted develop-YYYY-MM-DD) for the most recent successful develop pipeline (from GitLab). This job is scheduled to run on Sunday mornings.
  • Spack PR #37601 updates our GitLab CI rules to implement our new release strategy:
    • rebuild everything for release branches.
    • only copy previously built results for release tags.
    • Once this PR is merged we will be ready (from an infrastructure standpoint) to release Spack v0.20. At that point we will also start automatically creating rolling releases every weekend.

CI pipelines status

  • EC2-Other costs are down over the past few days thanks to improved network traffic managements (one NatGateway per-AZ and a new S3 Gateway). Next step: check if this results in faster downloads (and build times) in CI.
  • Earlier this week our intermediary signing key expired, causing some protected pipeline jobs to fail. This has been updated.
  • Jake is making progress on our new analytics dashboard (copy of GitLab + extra data). This will help us capture and dashboard more relevant metrics for our CI pipelines.

pcluster buildcache

  • Ryan is working with Stephen Sachs to publish a buildcache for pcluster. This involves creating a new runner image, karpenter provisioner, GitLab runner deployment, and Spack CI Stack. We expect an initial version of this buildcache to be available in the coming days.

Priorities

  • Get PR #37601 merged so we can cut release v0.20.
  • Populate the pcluster buildcache.
  • Deploy updated version of cache.spack.io and packages.spack.io.
  • Add a new pie chart to grafana showing CPU hours used per package.
Clone this wiki locally