Always

Open a PR to the GitHub MCO repository.

The teardown portion of the TestOnClusterBuildsRollout test fails thusly:

  utils.go:1097: Deleting machine ci-op-v5qcditr-46b3f-bh29c-worker-c-fcl9f / node ci-op-v5qcditr-46b3f-bh29c-worker-c-fcl9f
    utils.go:1098: 
            Error Trace:    /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:1098
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/onclusterbuild_test.go:103
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/helpers_test.go:149
                                        /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:79
                                        /usr/lib/golang/src/testing/testing.go:1150
                                        /usr/lib/golang/src/testing/testing.go:1328
                                        /usr/lib/golang/src/testing/testing.go:1570
            Error:          Received unexpected error:
                            exit status 1
            Test:           TestOnClusterBuildRollsOutImage
    utils.go:1097: Deleting machine ci-op-v5qcditr-46b3f-bh29c-worker-c-fcl9f / node ci-op-v5qcditr-46b3f-bh29c-worker-c-fcl9f
    utils.go:1098: 
            Error Trace:    /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:1098
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/onclusterbuild_test.go:103
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/helpers_test.go:149
                                        /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:79
                                        /usr/lib/golang/src/testing/testing.go:1150
                                        /usr/lib/golang/src/testing/testing.go:1328
                                        /usr/lib/golang/src/testing/testing.go:1312
                                        /usr/lib/golang/src/runtime/panic.go:522
                                        /usr/lib/golang/src/testing/testing.go:980
                                        /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:1098
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/onclusterbuild_test.go:103
                                        /go/src/github.com/openshift/machine-config-operator/test/e2e-layering/helpers_test.go:149
                                        /go/src/github.com/openshift/machine-config-operator/test/helpers/utils.go:79
                                        /usr/lib/golang/src/testing/testing.go:1150
                                        /usr/lib/golang/src/testing/testing.go:1328
                                        /usr/lib/golang/src/testing/testing.go:1570
            Error:          Received unexpected error:
                            exit status 1
            Test:           TestOnClusterBuildRollsOutImage

This part of the test should pass.

The way the test teardown process currently works is that it shells out to the oc command to delete the underlying Machine and Node. We delete the underlying machine and node so that the cloud provider will provision us a new one due to issues with opting out of on-cluster builds that have yet to be resolved.

At the time this test was written, it was implemented in this way to avoid having to vendor the Machine client and API into the MCO codebase which has since happened. I suspect the issue is that oc is failing in some way since we get an exit status 1 from where it is invoked. Now that the Machine client and API are vendored into the MCO codebase, it makes more sense for us to use those directly instead of shelling out to oc in order to do this since we would get more verbose error messages instead.