Hi Tom,
The admission controller mainly validates the cluster, although this also means that the admission controller will try to inject default values that may be missing from the cluster spec. We’ve tried to ensure that the default cluster values wouldn’t be dependent on having the admission controller running, which means the cluster should still start if you set :
deployments.admissionController=false.
However, if the cluster never starts without the admission controller (even with retrying), then I suspect the issue is some value isn’t being injected in time due to a race at startup. And if that’s the case I’ll need to see your chart values+overrides to look further.
What also seems telling is this timeout:
Timeout: request did not complete within requested timeout 30
Which can occur in cluster using many Persistent volumes or slow network pulling images. Try increasing pod-create timeout by adding the following to values file:
couchbaseOperator:
# name of the couchbase operator
name: "couchbase-operator"
....
commandArgs:
# Set to false if you are installing the CRD manually
create-crd: true
pod-create-timeout: 120