Wikimedia Release Engineering Team/Offsites/2018-12-Long Beach/Topics

Let's list all the things you want to talk about!

Topic proposals

Vote "+1" by adding your initials to the (interest: ) parenthetical

• Team Social Norms a la https://www.mediawiki.org/wiki/Wikimedia_Technical_Engagement/Team_Social_Norms (interest: gjg, ž, thc, dzd, liw)

• l10nupdate RFC: https://phabricator.wikimedia.org/T158360 (interest: gjg, )

• Zuul dependent pipeline https://phabricator.wikimedia.org/T94322 (interest: gjg, thc, anm)
  • see also: https://docs.google.com/document/d/1Sdqe5vCWu5UOVWIFyL_9vhzwMfwZzsRe2WP0uHn6pms/edit#

• CI data gathering, planning, and purposefulness (interest: gjg, thc, anm, dzd, liw)
  • tl;dr: how can we get better at being data driven with our CI changes without alienating our non-team contributors and/or be reactive to the needs of others in a timely manner?

• Zuul.v2 is no longer supported upstream (interest: thc, gjg, anm, dzd, liw)
  • we have effectively forked at this point, not ideal
  • Jenkins Gearman plugin unsupported
  • v3 vs jenkins
  • how to address jjb and zuul layout debt?
    • if we stick with jjb, should we write/enforce coding conventions and complementary linters?

• Jenkins "shifting-gears" (interest: thc, gjg, anm, dzd, liw)
  • https://jenkins.io/blog/2018/08/31/shifting-gears/
  • k8s jenkins vs breaking changes release vs nothing to support legacy installs?

(Antoine: Zuul v2 obsolete and Jenkins shifting gear are tightly coupled topic. It is all about: which software stack to use in the future).

• how do we manage the complexity of our CI/CD systems? (interest: liw)
  • probably ties into all the other topics

• zuul/layout.yaml debt (interest: thc, liw)
  • so many pipelines
  • so many templates
  • opportunities for consolidation/debt cleanup
  • how do we test pipelines/templates and CI in general?

• rethink test suite execution strategy (interest: dzd, liw)
  • we run all test suites for all package-manager/php/db flavors. should we?
  • what's the ideal test runner flow (a directed graph with parallel tasks) for satisfying coverage without redundancy? can quibble/jobs be refactored to implement it?
  • currently at 12 different variations. each new flavor increases load geometrically

• • random example: https://gerrit.wikimedia.org/r/#/c/mediawiki/extensions/Wikibase/+/465578/
    • Same tests, different extensions
    • Ran wmf-quibble-vendor-mysql-hhvm-docker (https://integration.wikimedia.org/ci/job/wmf-quibble-vendor-mysql-hhvm-docker/7817/consoleFull )

setup-docker-quibble-castor-load setup-docker-quibble-zuul-prepare setup-docker-quibble-submodules VisualEditor setup-docker-quibble-submodules Wikibase setup-docker-quibble-mw-install setup-quibble-npm-ls PHPUnit (without db) QUnit Selenium PHPUnit (with db)

• • • Ran quibble-vendor-mysql-hhvm-docker (https://integration.wikimedia.org/ci/job/quibble-vendor-mysql-hhvm-docker/20370/consoleFull )

setup-docker-quibble-castor-load setup-docker-quibble-zuul-prepare setup-docker-quibble-submodules Wikibase setup-docker-quibble-submodules VisualEditor setup-docker-quibble-mw-install setup-quibble-npm-ls PHPUnit (without db) QUnit Selenium PHPUnit (with db)

• • Antoine's Spreadsheet: https://docs.google.com/spreadsheets/d/1Fmcj3gwbUj9r0M8WI2W1lRjUC4XaznWZJ4kOeELmcgg/edit#gid=0
    • And patch: https://gerrit.wikimedia.org/r/#/c/integration/quibble/+/427099/

• Gerrit and releases in a post-chad world (interest: thc, gjg, mm, liw)
  • (Antoine: arent we supposed to hand it off to the mediawiki team?)

• Train automation goals planning (interest: thc, ž, gjg, anm, dzd, mm, liw)
  • One push deploy button!
  • Mukunda is making progress on scap swat

• scap vs https://pythonclock.org/ (interest: thc, mm, liw)

• Dev Productivity: Outcome 1: Local development is unified with testing and production (interest: gjg, ž, dzd, mm, thc, liw)
  • what are we doing here

• IRL Keysigning party! (if needed) (interest: liw)
  • within team and across teams?