This week, we have two new people starting in Release Engineering: Aki Sasaki (:aki) and Rok Garbas (:garbas). Please stop by #releng and say hi!

Modernize infrastructure:

This week, Jake and Mark added check_ami.py support to runner for our Windows 2008 instances running in Amazon. This is an important step towards parity with our Linux instances in that it allows our Windows instances to check when a newer AMI is available and terminate themselves to be re-created with the new image. Until now, we’ve need to manually refresh the whole pool to pick up changes, so this is a great step forward.

Also on the Windows virtualization front, Rob and Mark turned on puppetization of Windows 2008 golden AMIs this week. This particular change has taken a long time to make it to production, but it’s hard to overstate the importance of this development. Windows is definitely *not* designed to manage its configuration via puppet, but being able to use that same configuration system across both our POSIX and Windows systems will hopefully decrease the time required to update our reference platforms by substantially reducing the cognitive overhead required for configuration changes. Anyone who remembers our days using OPSI will hopefully agree.

Improve CI pipeline:

Ben landed a Balrog patch that implements JSONSchemas for Balrog Release objects. This will help ensure that data entering the system is more consistent and accurate, and allows humans and other systems that talk to Balrog to be more confident about the data they’ve constructed before they submit it.

Ben also enabled caching for the Balrog admin application. This dramatically reduces the database and network load it uses, which makes it faster, more efficient, and less prone to update races.

Release:

We’re currently on beta 3 for the Firefox 45. After all the earlier work to unhork gtk3 (see last week’s update), it’s good to see the process humming along.

A small number of stability issues have precipitated a dot release for Firefox 44. A Firefox 44.0.1 release is currently in progress.

Operational:

Kim implemented changes to consume SETA information for Android API 15+ data using data from API 11+ data until we have sufficient data for API 15+ test jobs. This reduced the number of high number of pending counts for the AWS instance types used by Android. (https://bugzil.la/1243877)

Coop (hey, that’s me!) did a long-overdue pass of platform support triage. Lots of bugs got closed out (30+), a handful actually got fixed, and a collection of Windows test failures got linked together under a root cause (thanks, philor!). Now all we need to do is find time to tackle the root cause!

See you next week!

http://coopcoopbware.tumblr.com/post/138749625280