GKE deployment about admiralty HOT 3 CLOSED

Answering myself, but if anyone get the issue:
if gke is configured as a routes based cluster, you'll get the error described above.
It works fine if gke is configured as a VPC native cluster.
Also, kubernetes >= 1.17 seems to be required to get v1.CSINode CRD support

from admiralty.

Hi @skalp ,

Proxy pods tolerate both taints since #44 was fixed, so you shouldn't have any problem with routes-based GKE clusters.

The pods deployed in a namespace with a multicluster.admiralty.io/elect: "" annotation set stay in Pending State.

The pods should be annotated with multicluster.admiralty.io/elect="", and the namespace should be labeled with multicluster-scheduler=enabled. It's a little convoluted, but that's how the webhook selects pods for now. We'll simplify soon.

Regarding CSINode support, please refer to #19. We actually just published a scheduler image that works with 1.16 (and maybe older versions). So you can override the image tag for the proxy-scheduler and candidate-scheduler deployments with 0.11.0-k1.16 and add , "--feature-gates", "CSINodeInfo=false"" to their args arrays.

The requirement for Kubernetes >=1.17 disappeared in the revamped documentation. I'll make sure it's clearer.

from admiralty.

@adrienjt Thanks for the clarifications

from admiralty.