February might be shorter than other months, but it wasn’t short on new things landing.
One of the benefits of a task execution platform is the ability to schedule tasks to run outside of the typical commit/build/test workflow. This can be done for rebuilding toolchains frequently to receive package updates or ship a product on a predefined interval.
TaskCluster’s hook service allows a developer to define a task and a schedule to run it on. However, this configuration is stored within the hooks manager, requires the developer to have the appropriate permissions within TaskCluster, and can create a secondary workflow for defining tasks to run on a repository.
Today, similar to how repositories could define a .taskcluster.yml as an entry point into scheduling tasks on push, support for a .cron.yml file has now been added to our Gecko branches that allows one to define tasks to run as well as a schedule.
While this work is very Gecko specific, and makes use of existing taskgraph generation logic, the concept can very much be reapplied to a repository hooked into TaskCluster, which leads me to the next amazing announcement.
You can read more about it here.
Github Integration Improvements
For some time now, github repositories under the mozilla org could define a .taskcluster.yml file in the root of the repository to schedule tasks for pushes and pull requests. Our integration service listens for events under the mozilla org using a webhook and then schedule the appropriate tasks based on that file.
We are fortunate, and thrilled!, to be able to participate in the Outreachy program and during this session, Irene worked hard with the team to deliver some awesome improvements to our github integration as well as supporting the new first class github support for integrations.
To read more about this, including checking out the new quickstart tool, head over to Dustin’s blog
Fast-azure blob storage
We also have Elena participating in the current Outreachy session improving our fast-azure-storage library and implementing it into some of our services. With this new version, the library supports the full blob storage API, allows specifying a schema for a blog entity, and adds the ability to use Shared Access Secrets for authentication.
I’m happy to report that Elena has shipped version 1.0! In addition to shipping the library, Elena has integrated a new API endpoint into taskcluster-auth that allows a shared access signature to be returned for access to a blob storage container.
Once tasks are scheduled for a push, a developer might find them in the position of wanting to reschedule that same task, maybe even many times. This is simple using Treeherder’s retrigger functionality, but it does not give the flexibility of specifying alternate parameters for the task and also relies on task graph duplication logic found in another integration service.
Rather than specifying actions within Treeherder or other tools that live outside of the in-tree task scheduling logic, a specification has been developed that describes how decision tasks can define actions during taskgraph generation, which tools, such as Treeherder, can use to assist a developer in performing actions on that task.
Beginning last year, we have been making incremental improvements to our front-end properties, such as our tools and documentation sites. While working on that it was clear that we needed better build tools and Neo was born. This got us far, but we wanted to make use of webpack and because TaskCluster’s core platform is built upon Node.js, we would like a way to build our applications with the same toolchain as our front-end.
Check out the project on github and help us make it even better!
As always you can find us in #taskcluster on irc.mozilla.org, or through one of our many projects on github.