Firefox/Input/Releases/1.8/PostMortem
From MozillaWiki
Contents
What Action was Performed?
- Update script
- push
- update vendor libs
- apache reload
- run crons
Issues before Release
- Stage was displaying a blank white screen 1 out of every 40 or 50 page loads.
- I didn't realize tests haven't been run since the 19th, on staging :-(
- Moved to Phoenix 2 weeks ago
- Jeremy Orem did a push on 6/21 thinking everything was fine. We had cancelled the push 1/2 an hour before. We had to quickly revert.
- Unsure about latest code commit. Belief was that everything was checked in and tagged.
- Fred was at Djangocon for a week
- Ryan was predisposed with emergency family stuff for a week and Socorro Work Week
- Dave was incredibly hammered with AMO
- Stephen was busy with Socorro, mozilla.com, mobile stuff for the past 2 weeks
Issues during Release
Production
- bug 600445 - Auto-Select default Version when changing Product
- bug 600448 - Mobile "--all--" option shouldn't include Firefox-desktop-build submissions
- bug 600627 - Input Production updated, but not to most recent Commit
Staging
- bug 600444 - Input staging/mobile staging are down for the count
- bug 595358 - Need DB dump from Input prod -> staging (I forgot that was open)
Perspectives
stephend
- Staging should have been up and should have been checked
- Created a release checklist
- No monitors: Nagios and Hudson and both would have caught this
ryansnyder
- lack of preparation
- end of quarter goals push tension, lack-of-focus and tiredness
davedash
- late pushes at the end of the day when people are unavailable
- push was pretty successful just with the issues when people were fresh
aakashd
- lack of a push checklist
- move up the start time
Solutions
- Move releases to 4pm to have availability of a dev who's 'fresh'
- Create a tailored release checklist
