Not being able to register CRD is not a fatal error

[redbaron]

Operator proceeds to checking if CRD is present and ready,
and if not, only then it is a fatal error.

[coveralls]

Coverage remained the same at 23.766% when pulling 46fdbb475cd45e695d43f5d052f78e52389f9a29 on redbaron:crd-no-fatal into c70905a on zalando-incubator:master.

[sdudoladov]

@redbaron at the absence of the operator CRD, at which point do you expect the operator pod to fail ?

[redbaron]

It fails on CRD wait status then , which remains Fatal error

time="2019-01-09T14:53:00Z" level=warning msg="Can't register customResourceDefinition operatorconfigurations.acid.zalan.do, will continue if it already exists" pkg=controller
time="2019-01-09T14:53:04Z" level=fatal msg="could not register Operator Configuration CustomResourceDefinition: customresourcedefinitions.apiextensions.k8s.io \"operatorconfigurations.acid.zalan.do\" not found" pkg=controller

[Jan-M]

But isn't the message confusing, why will it continue to fail if it already exists?

[redbaron]

But isn't the message confusing, why will it continue to fail if it already exists?

I don't understand. If CRD already exists, then irrespectively of CRD creation it moves to checking CRD status and only if CRD status check fails, operator fails too.

[redbaron]

In other words, it fails on what is important : CRD is not registered in the cluster

[sdudoladov]

hey @redbaron :) any updates on this PR ?

[FxKu]

Thanks for the update. With the new update mechanism there are also no confusing message anymore.

[FxKu]

👍

comments addressed in a subsequent commit

[sdudoladov]

👍

[sdudoladov]

@redbaron thanks for the contribution 👍