Change apiVersion to zalando.org/v1

[mikkeloscar]

This changes the apiVersion defined for the AWSIAMRole CRD from amazonaws.com/v1 to zalando.org/v1.
This is unfortunately a breaking change that we feel the need to make as we are preparing to roll this out in Zalando.

The reason to make this change is to ensure that Zalando as a company is not breaking any Trademark laws. While this project has been developed by me until now. As I'm an employee of Zalando and because of European Copyright laws, the copyright of the project falls to Zalando as this is related to my daily work. Therefore it makes sense to use zalando.org/v1 instead of coming up with another name which may be trademarked by someone else.

@seh already brought up this issue in the original PR.

Migration steps

1. Apply new CRD:

   kubectl apply -f docs/aws_iam_role_crd.yaml
   

2. Copy old resources to new version:

   #!/bin/bash
   
   for ns in $(kubectl get ns --no-headers -o "custom-columns=:metadata.name"); do
       for res in $(kubectl --namespace "$ns" get awsiamroles.amazonaws.com --no-headers -o custom-columns=:metadata.name); do
           kubectl --namespace "$ns" get awsiamroles.amazonaws.com "$res" -o json | sed 's#amazonaws.com/v1#zalando.org/v1#' | kubectl apply -f -
       done
   done

3. Deploy new version of controller >=0.1.0 (See https://github.com/mikkeloscar/kube-aws-iam-controller#setup).

4. Delete old resources and old CRD:

   kubectl delete crd awsiamroles.amazonaws.com
   

5. If you are using RBAC, you also need to update the ClusterRole to have zalando.org instead of amazonaws.com as apiGroups.

[coveralls]

Pull Request Test Coverage Report for Build 112

• 4 of 4 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 67.454%

---

Totals
Change from base Build 109:	0.0%
Covered Lines:	514
Relevant Lines:	762

---

💛 - Coveralls

[seh]

Given the version bump, do you have any plans to phase out the older IAM role binding technique of mounting Secrets named with a special prefix? I'd prefer to be able to deploy the controller at least with the pod cache disabled. If we can't remove it for fear of breaking existing users, perhaps we could disable it via command-line flag.

I was hoping that the next major version bump would be our chance to remove that feature.

[mikkeloscar]

I created #25 to track this. I agree we should phase it out.

[Shinzu]

@mikkeloscar if your deployment is with rbac you must also migrate the clusterrole

[mikkeloscar]

@Shinzu thanks, added this as a step in the description above!