Deployments/Features Process/E2 Notes

From Wikitech


  • if anything new to be deploy, matthias deploys
  • weekly slots every tuesday
  • if small changes that aren't urgent we wait until platform
  • every change is pushed to gerrit, then to master then deployed through platform or the aft window, or sometimes with kaldari's deploy :)
  • prototype branch
    • not gated
    • eeprototype.wmflabs
    • works out decently well
    • time consuming to merge back to master, keeping in sync
    • betacluster could be useful if that was able to pull in a certain branch/extension
      • only master is visible on betalabs
      • maybe worth a conversation

kaldari, benny on Echo

  • fairly similar
  • if two live versions,
    • we each take one for deploy
  • not a strict schedule
  • not using separate branches
  • they keep eeprototype on master
  • try pushing to master often, less than a week
  • core code that sits and rots -- important
  • usually ok with non-performance
  • need a strong db type person who can review on core changes
  • does get good updates from Chris
    • Chris is pretty good proactive wise
  • for performance they don't have anyone
  • terry's approach is to push the envelope and push yourself if you have to, if they complain then ask them to review :)
  • everything is configurable
  • ResourceLoader has a hash update of css/js, gets stuck on old versions when echo pushes out, fatal problems
    • go on fenari, touch, resync