Wikimedia Release Engineering Team/Deployment pipeline/2018-11-08
Appearance
2018-11-08
[edit]Last Time
[edit]General
[edit]- process suggestion
- alex: run the test variant prior to building production docker images
- joe: builds the test variant in first phase to include software for unit tests, but don't publish images until people vote +2 on code review, blubber image builds are very fast due to the cache
- lars: building docker image locally for testing is a good plan
- joe: deployment to production or deployment to staging -- currently is scap helm -- but we need to figure out authorization
- alex: removing release engineering requires a large culture change for service owners, particular for MediaWiki where releng acts as an intermediate
- fabian: how does rollback work with this?
- joe: scap helm deploys to canaries
- Versioning of docker images
- alex: current zotero version was hard to find -- we need another way to do that
- joe: someone should create a proposal
- thcipriani: (stupidly :)) I'll do that - TODO create task for versioning images, post to #wikimedia-pipeline, come up with proposal based on comments
RelEng
[edit]- contint1001:/var/lib/docker
- Currently has CI images, pipeline images are cleaned up after they're built, but it seems easy to fill the root partition
- Not sure what the options are for adding space to that machine? /srv is large, but it is for zuul merger, so ideally wouldn't use that
- Adopt JSON as blubber's internal configuration format
- Lars working on this
SREs
[edit]Services
[edit]- Marko's Email
- Zotero, Node10 event platform, Node6 LTS EOL
- Zotero helm chart still needs review it seems https://gerrit.wikimedia.org/r/#/c/operations/deployment-charts/+/466287/