Wikimedia Release Engineering Team/CI Futures WG/Meetings/2019-02-26

Kickstart meeting 2019-02-26

On Google Hangouts Meet.

• Brennen
• Lars
• Zeljko

Agenda

The goal of this meeting is to get started, get organised, and have a plan for how to achieve the goal of the working group.

• Are the participants OK participating in the work group?
  • Brennen: OK
  • Lars: OK.
  • Zeljko: OK

• What is and isn't in scope of the WG?
  • Evaualute current CI tools (Zuul, Gearman, Jenkins), and possible replacements.
  • Code hosting, code review probably to stay with Gerrit.
    • Brennen: I think I'd like to see this up for discussion eventually, but probably out of scope here.
  • Ticketing probably to stay with Phabricator.
  • In-scope: everything after a change is pushed to Gerrit until software runs in production.
  • Given/assumed:
    • production will run services/sites on containers in Kubernetes
    • builds will ideally do that too

• Suggestion for plan:
  • Gather requirements
  • Evaluate candidates
  • Discuss, come up with a recommendation
  • Write and publish (short) report
    • What did we consider? What are requirements?
    • Rationale for recommendation.
  • TODO: Lars to check with Greg how much time we should spend on this

• This week we interview people privately to collect requirements and agree on evaluation method (spreadsheet? reports?).
  • Lars has already inteviewed Greg, Tyler, and asked Jeena about any CI related feedback from the developer interviews.
  • Interview with Antoine scheduled for Wednesday.
  • Anyone else in RelEng or otherwise?
  • Wiki page with results: https://wikitech.wikimedia.org/wiki/Deployment_pipeline#Requirements_for_pipeline_tooling
  • TODO: Z: req: debugging CI (when local works) is super-annoying, this should be made easier
  • TODO: Z: req: we need to store at much artifacts as possible, when a test fails, VM or container should be made available for debugging to the dev

• Discuss current list of requirements and candidates.
  • Postpone this for a later meeting and offline discussions.
  • Z: Requirements list potentially a good use for voting.

• This week also start collecting list of possible candidates.
  • Same page as above.
  • Current: Zuul v2.5, Gearman, Jenkins; Antoine says this is untenable in the long run
  • Zuul v3
  • GitLab-CI (not for git hosting?)
    • B: One advantage here would be as a wedge for introducing GitLab itself, if that's worth exploring
  • Go CD
  • Buildbot
  • Jenkins on its own
  • possible list: https://en.wikipedia.org/wiki/Comparison_of_continuous_integration_software
  • others???
    • B: I'm a little curious about https://sourcehut.org/ (probably not, but maybe worth a glance)
    • Circle-CI (license?)
    • https://semaphoreci.com (license?)

• TODO: Lars to make a weekly blog post with updates of the WG work, for publicity
• TODO: Lars to ask publically or within WMF for suggestions for candidates
• TODO: Ask for suggestions within our personal networks

• Evaluation method? Once we have a list of requirements and a list of candidates, how do we evaluate them?
  • We should have a short list of very hard requirements
  • Serious candidates need a demo / description of the workflow
    • Probably sufficient to build a git repo - if it works with git, we're probably good
  • A huge spreadsheet?
  • Short one-page reports on each candidate?
    • Why is (or is not) each candidate an acceptable choice?
  • How do we divide the evaluation work?
    • Have at least two people look at each thing?

• TODO: Lars to set up a new meeting for Thursday. One hour before the Docker-SIG meeting.
• TODO: Lars to create a Super Hard Requirements section
• TODO: Lars to move list to subpage or its own page

For the future

• Ask Dan if he wants to join the WG as well.
• Finish list of requirements
• Divide candidates for evaluation
• Add list of very hard requirements to Deployment pipeline page