Panic on startup if apiserver is unreachable

[ryanfaircloth]

humio-operator/images/helper/main.go

Line 428 in da5e41a

func initMode() {

In some cluster deployments a condition can occur where an init container connecting to the api server very quickly can receive a connection error. "Connection refused" and "500" errors should be retry-able for a configurable amount of time, default to 60s with 2s sleep

[SaaldjorMike]

I guess the main benefit here would be to retry things based on our own retry logic vs having to rely on container restarts which would also retry those requests, but overall I'm not sure this is super critical. It is obviously "nicer" to retry and not panic, but if you have a k8s setup with a ton of connection issues or so when talking to the api server, then its probably more important to fix those issues.

[ryanfaircloth]

this actually came up in EKS where aws handles scale of the control plane. Sometimes its just transient network issues other times it could be a delay in the cni