Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign upTaskcluster: use Windows instances is multiple AWS regions #23652
Labels
Comments
|
Docs for In the meantime, I was pointed at the source for the exact set of API and their schema: |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The CI queue is slow right now because Windows tasks for the PR being tested keep being killed and restarted.
https://tools.taskcluster.net/groups/UowzfhVIR7q_1bl9hagSeQ/tasks/HA3OnWLmS-yP0wDfJiZyLA/runs/3/logs/public%2Flogs%2Flive.log
The above is the fourth run of this particular task.
https://tools.taskcluster.net/aws-provisioner/servo-win2016/errors shows many errors with code
Server.SpotInstanceTerminationandInsufficientInstanceCapacity.People on #taskcluster on IRC suggest that this problem is made worse by our worker type configuration at https://tools.taskcluster.net/aws-provisioner/servo-win2016/edit only specifying the
us-west-2region. With our usage spread over multiple AWS regions, we would be less subject to low availability in a given region.However, AMIs are region-local and we manage the AMI for Windows CI ourselves, and it’d be nice to not increase the number of steps needed every time we want to update it.
Possible steps to fix this:
etc/taskcluster/windows/build-ami.py, after the AMI is built:taskcluster api awsProvisioner workerType servo-win2016-stagingto get the current definition of theservo-win2016-stagingworker typeImageIdfor the new AMItaskcluster api awsProvisioner updateWorkerType servo-win2016-stagingtaskcluster api secrets setso that manual copy-paste into 1Password is not needed.)servo-win2016-stagingworker type.ImageIdfrom theservo-win2016-stagingdefinition, copies the AMI into multiple regions, and updatesservo-win2016to use them.