It will be nice to allow to specify labels for node selection for GKE integration.
It will allow to target node pools with SSDs, GPUs, etc.
gke_container:
image: gradle:jdk8
cluster_name: cirrus-ci-cluster
zone: us-central1-a
node_selector_terms:
- key: "kubernetes.io/hostname"
operator: "In"
values: "gke-test-cluster-default-pool-926ddf80-f166"
Weird enough GKE doesn't have labels to indicate SSD availability to cover @ccavolt case from #264. Raised the issue on StackOverflow where GCP support team answers such questions much faster then by email 馃槄
Just stumbled upon a case where this would be beneficial.
If you get the chance to work on this please at the same also add support for taints and tolerations (in some cases they go in-hand with affinity):
https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
Also I would recommend to keep this whole section (affinity and tolerations) exactly like in kubernetes, i.e. camelCase instead of snake_case which saves users from heaving to learn another syntax and saves you from documenting everything in detail (you can just refer to the k8s docs for those sections).
Most helpful comment
Just stumbled upon a case where this would be beneficial.
If you get the chance to work on this please at the same also add support for taints and tolerations (in some cases they go in-hand with affinity):
https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
Also I would recommend to keep this whole section (affinity and tolerations) exactly like in kubernetes, i.e.
camelCaseinstead ofsnake_casewhich saves users from heaving to learn another syntax and saves you from documenting everything in detail (you can just refer to the k8s docs for those sections).