CI: 2023 05 05
Zack Galbreath edited this page May 5, 2023
·
1 revision
- Aashish Chaudhary
- Alec Scott
- Dan LaManna
- Jacob Nesbitt
- Massimiliano Culpo
- Scott Wittenburg
- Tamara Grimmett
- Todd Gamblin
- Zack Galbreath
- Dan presented a demo that's looking really nice!
- Next step is to deploy it, either overwriting the existing site or making a new
-dev
one. - Related: we should include size in the index.json to avoid a requiring S3 access to build this site.
- We should also investigate similarly refreshing packages.spack.io as time allows so these two companion sites have a similar look & feel
- Our cluster upgrade last week did not go as smoothly as we would have liked, but we are now at a point where all our services are operational again. We will keep an eye on our dashboards to make sure we're not experiencing a higher-than-usual failure rate from our GitLab CI pipelines.
- That being said, the cluster migration did complete successfully, so we're hoping to see reduced data transfer costs in the coming weeks.
- Jake is making progress with new prometheus queries that will allow us to better capture cost-per-job and per-package stats.
- Sealed Secrets PR has been merged and deployed. It is working as intended.
- Scott drafted a design document outlining how we will implement this new type of release + snapshot buildcache.
- Some decisions that we made during this meeting:
- Create new snapshot releases on a weekly basis
- Use the same strategy for snapshot mirrors and point release: copy existing packages from the develop/release buildcaches.
- Use the lock file from the successful develop pipeline to determine what packages should be included in the snapshot buildcache.
- Keep snapshots for 6 months
- Alec & Todd are going to discuss how to smoothly handle upgrading from one snapshot to the next
- Removed inadvertent dependency on trilinos@develop from E4S stack
- Install a new cronjob to prune the develop buildcache.
- Implement and deploy our new strategy for release pipelines.
- Push a live public demo for the redesign of cache.spack.io.
- Keep an eye on costs & pipeline success rate. Respond swiftly if success rates drop below 97%.
- Keep working towards on costs-per-job & better per-package metrics.