CI: 2023 11 03
Zack Galbreath edited this page Nov 3, 2023
·
1 revision
- We updated our analytics database with more useful columns, such as package name, version, and compiler name.
- We are working on transitioning our job failure error taxonomy data from OpenSearch to the Analytics DB.
- Analytics database retention policy: let's keep data for six months, and dump older data to S3 rather than deleting it outright.
- We modified our Karpenter provisioners to apply a custom tain to our pipeline builder nodes. This prevents other (non-pipeline) pods from accidentally getting scheduled on them.
- We started working on storing our GitLab CI jobs in OpenSearch. This will make it easier to search through job output.
- We are investigating whether switching to ninja by default will increase the performance of our CI pipelines.
- We continue to work on content addressable tarballs for our binary mirrors.
- Assist the spack team as they prepare for the upcoming v0.21 release
- Investigate whether we can enable "login with GitHub" for Analytics database access
- Investigate API token access for prometheus.spack.io
- Draft a proposal for how to future-proof binary mirror layout formats
- Consider switching to zstd compression at the same time