This past week, the release engineering managers – Amy, catlee, and coop (hey, that’s me!) – were in Washington, D.C. meeting up with the other managers who make up the platform operations management team. Our goal was to try to improve the ways we plan, prioritize, and cooperate. I thought it was quite productive. I’ll have more to say about next week once I gather my thoughts a little more.

Everyone else was *very* busy while we were away. Details are below.

Modernize infrastructure:

Dustin deployed a change to the TaskCluster authorization service to support “named temporary credentials”. With this change, credentials can come with a unique name, allowing better identification, logging, and auditing. This is a part of Dustin’s work to implement “TaskCluster Login v3” which should provide a smoother and more flexible way to connect to TaskCluster and create credentials for all of the other tasks you need to perform.

Windows 10 in the cloud is being tested. All the ground work is done to make golden AMIs, mirroring the first stages of work done for Windows 7 in the cloud. Being able to perform some subset of Windows 10 testing in the cloud should allow us to purchase less hardware than we had originally anticipated for this quarter.

Improve CI pipeline:

One of the subjects discussed at Mozlando was improving the overall integration of localization (l10n) builds with our continuous integration (CI) system. Mike fixed an l10n packaging bug this week that I first remember looking at over 4 years ago. This fix allows us to properly test l10n packaging of Mac builds in a release configuration on check-in, thereby avoiding possible headaches later in the release cycle. (https://bugzil.la/700997)

Armen, Joel, Dustin, and Greg worked together to green up even more Linux test jobs in TaskCluster. Among other things, this involved upgrading to the latest Docker (1.10.0) and diagnosing some test runner scripts which use 1.3GB of RAM – not counting the Firefox binaries they run! This project has already been a long slog, but we are constantly making progress and will soon have all jobs in-tree at Tier 2.

Release:

Ben and Nick started designing a new Balrog feature that will make it possible to change update rules in response to certain events. Ben is planning to blog about this is more detail next week.

It’s was a busy week for releases. Many were shipped or are still in-flight:

• Firefox/Fennec 45.0b3
• Firefox/Fennec 45.0b4
• Firefox 44.0.1, quickly followed by Firefox/Fennec 44.0.2 due to a security issue
• Firefox 38.6.1esr
• Thunderbird_45.0b1 was shipped to Windows and Mac populations. We are planning a follow-up 45.0b2 in the near future to pick up the swap back to gtk2 for Linux users.
• Firefox 45.0b5 (in progress)
• Thunderbird 38.6.0 (in progress)

As always, you can find more specific release details in our post-mortem minutes: https://wiki.mozilla.org/Releases:Release_Post_Mortem:2016-02-10 and https://wiki.mozilla.org/Releases:Release_Post_Mortem:2016-02-17

Operational:

Kim landed a patch to enable Mac OS X 10.10.5 testing on try by default and disable 10.6 testing. This allowed us to disable some old r5 machines and install around 30 new 10.10.5 machines and enable them in production. Hooray for increased capacity! (https://bugzil.la/1239731)

See you next week!

http://coopcoopbware.tumblr.com/post/139190778845