Issues when deploying Autopilot/multiple clusters
As part of testing for https://gitlab.developers.cam.ac.uk/uis/devops/raven/infrastructure/-/issues/136 I've found two issues which need to be resolved. They are:
-
Currently it is not possible to deploy two clusters to the same project due to the recent addition of the no scale down alert. This alert needs to be tweaked so that it is specific to the deployed cluster and the name of the resources are unique.
-
There is currently an open issue for the terraform google provider
>4.3.0
whereby deployment of an Autopilot enabled cluster fails with the following error.Error: googleapi: Error 400: Max pods constraint on node pools for Autopilot clusters should be 32
The workaround for this at the moment is either lower the google provider to 4.3.0 or implement an empty
ip_allocation_policy {}
block for the cluster resource. I think the second option is probably better as the issue has been open for over a year with no resolution so we don't really want to restrict our provider to 4.3.0 indefinately.