tl;dr: 

Taskcluster, the platform supporting Firefox CI, will be moving to a new 
hosting environment during the tree closing window (TCW) this coming Saturday, 
Nov 9. Trees will be closed from 14:00 UTC to 23:00 UTC. CI services will be 
available as soon as possible thereafter, pending verification of the new setup.

More verbose:

While originally scheduled for September, the Taskcluster hosting migration was 
pushed back to the upcoming TCW in November to allow more bake time for the new 
worker provisioning system, but also to give community projects more time to 
prepare for migration.

The primary driver for this migration was a separation of concerns between 
Firefox CI and all the other various Mozilla projects that also require CI. 
After this migration, there will be two production Taskcluster deployments 
where right now there is one:

1. Firefox CI - This cluster needs to be secure and optimized for the release 
process of Firefox and associated products (mobile browsers, etc). We have 
multiple teams responsible for shepherding Firefox code through the release 
process, and each team will be responsible for a different aspect of the 
Firefox CI deployment:

  Deploying and running Taskcluster services  - CloudOps
  Worker management and imaging               - RelOps
  Cluster admin (granting scopes, roles, etc) - RelEng
  Software issues with Taskcluster itself     - Taskcluster team

2. Community projects - These projects might be small one-offs or might be 
speculative endeavors that may eventually become products of their own. 
Generally these projects are running on github. There is a greater capacity and 
tolerance for experimentation in this deployment, and projects are able to 
self-administer. The breakdown of responsibilities for the Community cluster is 
as follows:

  Deploying and running Taskcluster services  - CloudOps
  Worker management and imaging               - Taskcluster team, or individual 
projects
  Cluster admin (granting scopes, roles, etc) - Taskcluster team, or individual 
projects
  Software issues with Taskcluster itself     - Taskcluster team

As a result of this migration, root URLs will be different for each cluster:

Firefox CI: https://firefox-ci-tc.services.mozilla.com
Community:  https://community-tc.services.mozilla.com

Tools like treeherder and in-tree references will be updated to use these new 
URLs automatically as part of the downtime. Existing URLs, e.g. links to 
artifacts, will continue to work once the current monolithic deployment is put 
into read-only mode on Monday, Nov 11.

If you notice anything broken after the weekend, please file a bug. Here is a 
list of the best Product::Components to use in Bugzilla, depending on what is 
broken:

  Deploying and running Taskcluster services - Cloud Services::Operations: 
Taskcluster
  Worker management and imaging              - Taskcluster::Workers
  Cluster admin                              - Release Engineering::Firefox-CI 
Administration
  Software issues with Taskcluster itself    - Taskcluster::Services
  Anything else                              - Taskcluster::General

Please be sure to indicate whether you're on the Firefox CI or Community 
deployment.

The bug tracking this work is here: 
https://bugzilla.mozilla.org/show_bug.cgi?id=1546801 

You can find the Taskcluster team on Slack or IRC in #taskcluster if you have 
concerns or questions.

cheers,
--
coop


On Tuesday, August 6, 2019 at 2:48:36 PM UTC-4, Dustin Mitchell wrote:
> *tl;dr:* Taskcluster, the platform supporting Firefox CI, will be moving to
> a new hosting environment during the tree-closing window at the end of
> September.  This is a major change and upgrade to the platform, but may
> cause some bumps along the way.
> 
> Taskcluster is in the midst of a few migrations [1], and one of those is
> moving to a new hosting environment.  This has a few advantages:
>  * Firefox CI will be handled by dedicated teams, separate from the TC team:
>    * Service operations will be managed by cloudops
>    * Administration will be managed by releng
>    * Workers will be managed by relops
>  * Hosting will be in a more trusted environment (GCP instead of Heroku)
>  * The services will have a new, faster UI
>  * Worker provisioning will be faster and more reliable, and cover more
> cloud environments
>  * ..lots of other achievements unlocked by shedding 5+ years of legacy
> 
> The downside is that URLs will be changing, likely leaving some dangling
> pointers.  We've anticipated some of the most common (such as
> https://queue.taskcluster.net/ URLs) and are planning workarounds, but
> nonetheless anticipate some dead links for a while following the switch.
> 
> This work is tracked in bug 1546801 [2].  If you have questions or
> concerns, please get in touch with me or anyone on the team, or file a bug
> in Taskcluster :: General and we will triage it appropriately.
> 
> Dustin
> 
> [1]
> https://mana.mozilla.org/wiki/pages/viewpage.action?spaceKey=TAS&title=Migrations
> [2] https://bugzilla.mozilla.org/show_bug.cgi?id=1546801

_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform

Reply via email to