[ACTION REQUIRED] Migrate your CI jobs to Community Clusters

[Davanum Srinivas]

Hi Folks, 

SIG K8s Infra + SIG Testing are trying to finish moving jobs out of the old cluster: default Google-internal GKE cluster. We're asking the broader community to help migrate your CI jobs.Why?

• The old default cluster has been around since v1.6 with tons of old CI jobs and does not have guaranteed QOS, so workloads are less stable. The replacement community owned clusters enforce guaranteed memory/CPU QOS and CI should be more stable there.
• We depend on go.k8s.io/oncall test-infra-oncall entirely for access and managing this cluster. The k8s infra clusters can potentially be managed by any community member.
• We're attempting to split the CI costs between the existing GCP credits and the new AWS credits. We have a lot of unused AWS capacity and we also want to ensure that as many jobs as possible work flexibly across clouds.

How?

1. Identify your CI jobs that do not depend on external cloud assets, for example unit test, linting, build. Jobs that e.g. create external GCP e2e clusters are not good candidates at this time as we're focusing on migrating
2. Add cluster: eks-prow-build-cluster to the top level Prow Job spec.
3. Revert the cluster change if you see CI failures. These clusters are not quite identical but in theory many CI workloads will "just work".

You can filter prow.k8s.io by the CI cluster like https://prow.k8s.io/?cluster=defaultThanks for helping us make CI more sustainable,

For more information, please see https://github.com/kubernetes/test-infra/issues/29722  

 – SIG Testing and SIG K8s Infra

--

Davanum Srinivas :: https://twitter.com/dims