Wikimedia Release Engineering Team/Checkin archive/20150901



Team Business[edit]

  • FYI
    • Aaron's multi-DC work in public RFC meeting, Wednesday Sept 2, 21:00 UTC

Deploy tooling[edit]

Objective: Reduce number of deploy tools from 3 to 2 Key result: Migrate all Service team owned services and MW deploys to scap3 - task T109926

We don't have a KPI to track this one, but it's easy to measure success ("did we retire trebuchet and ansible?").

Open questions: 1) Still doable (retirement of Trebuchet and Ansible by December)? 2) Request: We probably want specific subtasks of that one for the individual Service Team owned services 3) Note: The current "scap3" sprint does not need to be tracked by this task (T109926) because it's our Q1 work.

Migrate Gerrit to Differential[edit]

Objective: Retire Gerrit in favor of Phabricator (Differential) Key results: .... tbd ...

Open questions: 1) We need to own the Gerrit-Migration board and make it reflect reality (what actually needs to happen). I presume that will take a conversation between at least Chad, Mukunda, and myself (and Quim? Andre?). 2) I've filed a meta-task to track this work (the creation of the plan):

TODO: Meeting of Roadmap doom: Greg (optional), Chad, Mukunda, Antoine? TODO: Greg make the business case stuff

  • get rid of all the various glue bots

NB: Keep in mind that we've allocated ourselves two quarters to complete this work.

CI Scaling[edit]

Objective: Reduce CI wait time Key result: CI cluster responds to spike in queued builds by starting and registering additional jenkins slaves

We can use the "Jenkins/Zuul queue wait" KPI to track the effectiveness of this work: .

Open questions: 1) Is this task still the right task to judge the completion of this work?

2) Are the blockers of that task still accurate? IOW: when all of those blockers are completed we can consider the work done (whether or not it makes a change to the KPI)? 3) Doable in 3 months?

Zuul gate time KPI attempt

On going:

   ** build on a labs instance (integration-dev)
   ** copy .qcow2 image to /var/www/html
   ** curl from labnodepool
   ** sudo -H -u nodepool -s
   ** cd && . .profile
   ** openstack image create .....


  • Refactor MediaWiki tests. Split unit tests in their own jobs and speed up the lame 'integration' tests (10 minutes with Zend).
  • figure out a solution to cache npm/pip/composer/rubygems modules (tarballs and compiled)

Q1 goal: nodepool infra build. at least 1 production grade job using Q2: CI cluster responds to spike in queued builds by starting and registering additional jenkins slaves (and migrate more jobs) Q3: migrate rest / phase out legacy


Scrum of Scrums[edit]
Blocked on us:

Isolated CI instances CI Scaling[edit]
Quarterly Priority: Disposable VMs -

Deployment Cabal[edit]

Developer Tooling (MW-Vagrant, MW-Selenium, etc.)[edit]


Beta Cluster[edit]

Other Work[edit]


Please add your time off to your gcal, **Phabricator**, and ADP, as appropriate

  • Chad - Sept 7-11 (last minute vacation. mostly reachable by e-mail) Sept 18th & 28th (Music festivals/shows)
  • Željko planned to be offline on Wednesday September 2 but that has 1% chance of happening, sick kid
  • Monday Sept 7th - US Holiday (Labor Day)
  • Tyler: Sept 8th—in mountains
  • Mukunda: Sept 4th (This friday) (I can attend the sprint meeting, taking the afternoon off)
  • Andrew out this Friday