Skip runtimeclass test unless provider GCE

[bobbypage]

What type of PR is this?

/kind bug
/kind cleanup

What this PR does / why we need it:

This test case requires special test-handler setup which is only done for gce clusters created by kube-up scripts. Let's skip the test when run under other providers since it will fail under other environments due to test-handler not being setup.

Which issue(s) this PR fixes:

Fixes #

Special notes for your reviewer:

See #81915 (comment) for some old discussion about this.

Does this PR introduce a user-facing change?

NONE

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

[bobbypage]

/cc @jprzychodzen

[SergeyKanzhelev]

/triage accepted
/priority backlog
/assign

[jprzychodzen]

/retest

[jkaniuk]

I think that should run a Pod requesting a RuntimeClass with scheduling without taints should also be skipped on GKE

[bobbypage]

I think that should run a Pod requesting a RuntimeClass with scheduling without taints should also be skipped on GKE

Thanks, also added skip for this test. It also required runtime class handler to be setup.

[bobbypage]

/assign @SergeyKanzhelev

[jkaniuk]

/retest

[jkaniuk]

/retest

[jkaniuk]

/lgtm
/retest

[k8s-ci-robot]

@bobbypage: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name	Commit	Details	Required	Rerun command
pull-kubernetes-node-kubelet-serial	4b056a6	link	false	/test pull-kubernetes-node-kubelet-serial

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[ehashman]

These tests are broken/disabled on OpenShift right now, probably for this exact reason: https://bugzilla.redhat.com/show_bug.cgi?id=1923737#c5

I have one concern though... this feature is supposedly GA: kubernetes/enhancements#585 (comment)

Hence I don't think we should have environment-specific tests for it. Before we merge, what other coverage do we have for RuntimeClasses?

[SergeyKanzhelev]

Hence I don't think we should have environment-specific tests for it.

It is configuration specific, I think environment check here is a simplest proxy to validate that the needed runtime class was configured for the test.

Before we merge, what other coverage do we have for RuntimeClasses?

I don't think there are any coverage for success path execution fro the pod scheduling with the runtime class. Reject is being tested all right.

[ehashman]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: bobbypage, ehashman

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• test/e2e/common/OWNERS [ehashman]
• test/e2e/node/OWNERS [ehashman]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[liggitt]

didn't this skip tests that previously ran successfully in other envs (like https://testgrid.k8s.io/sig-release-1.22-blocking#kind-1.22-parallel&include-filter-by-regex=requesting%20a%20RuntimeClass)?

shouldn't we detect whether the runtimeclass is loaded rather than pinning to a particular provider?

[bobbypage]

didn't this skip tests that previously ran successfully in other envs (like https://testgrid.k8s.io/sig-release-1.22-blocking#kind-1.22-parallel&include-filter-by-regex=requesting%20a%20RuntimeClass)?

shouldn't we detect whether the runtimeclass is loaded rather than pinning to a particular provider?

Thanks Jordan for mentioning that. I wasn't aware that kind also sets up runtime class handler, it looks like it does so here.

shouldn't we detect whether the runtimeclass is loaded rather than pinning to a particular provider?

I agree this would be ideal, but unfortunately we don't really know easily during runtime if the runtime class is configured because it requires the underlying CRI implementation (containerd / crio) config to have explicit test-handler configured. AFAIU I'm aware this is only setup for kind and GCE provider. I suppose we could do something like have the test SSH into the node and dump the containerd config file and grep for test-handler but it seems a bit hacky as well :)

[hakman]

@bobbypage @liggitt These tests are used by other projects too (like kOps - kubernetes/kops#12974).
https://prow.k8s.io/view/gs/kubernetes-jenkins/pr-logs/pull/kops/12974/pull-kops-e2e-k8s-gce-ci/1471056821153173504

Usually tests that need a special setup to run are named [Feature: and can be later skipped by the [Feature:.+] pattern. Would this be acceptable in this case too?