KubernetesServer in crud mode should fill out missing metadata

[shawkins]

An object created by Kubernetes server will not have additional metadata fields set such as uid, namespace, resourceVersion, and generation. The workaround is to specify those on the object being created - but this is not inline with expected behavior. Frameworks like the java operator sdk are making assumptions about those values being non-null.

This is somewhat related to #2802

[rohanKanojia]

@shawkins : Thanks a lot for reporting this. I agree we should fill out these fields. Do you have some approach in mind on how to tackle this? Maybe we can fill out these fields before persisting fields in MockServer map:

kubernetes-client/kubernetes-server-mock/src/main/java/io/fabric8/kubernetes/client/server/mock/KubernetesCrudDispatcher.java

Lines 83 to 89 in 6e439a8

	switch (method.toUpperCase()) {
	case POST:
	return handleCreate(path, request.getBody().readUtf8());
	case PUT:
	return handleReplace(path, request.getBody().readUtf8());
	case PATCH:
	return handlePatch(path, request.getBody().readUtf8());

[shawkins]

It seems a little involved with the current logic in KubernetesCrudDispatcher. You need to update both the AttributeSet and string value of the object. And to glean the namespace you need access to AttributeSet values. This would look something like:

shawkins@36bf842

[shawkins]

@manusa @rohanKanojia opened #2943 and fabric8io/mockwebserver#59

[manusa]

Hi @shawkins
We'll check your PR. However it involves releasing MockWebServer which we don't have an ETA yet.

[shawkins]

@manusa do you want to back out the pom change and using the namespace from the path for now?

[manusa]

It will certainly speed up things. If we cut a release for MockServer we should make the most of the effort by at least upgrade dependencies and perform some other checks which will inevitably take some time.

[shawkins]

Pulled the mockwebserver upgrade out.