K8s 1.11 fails on generateName when upgrading pipelines

[timbuktuu]

When attempting to upgrade Kubeflow Pipelines by kubectl apply -f deploy.yaml leverages a k8s deployment job with generateName. This works on initial Kubeflow Pipelines cluster creation as one calls kubectl create -f, but fails when upgrading via kubectl apply -f in Kubernetes 1.11. Backing the cluster down to 1.10 allows to work as expected.

'kubectl apply -f https://storage.googleapis.com/ml-pipeline/release/$PIPELINE_VERSION/bootstrapper.yaml"

Similar issues are being reportes elsewhere.

[vicaire]

@IronPan, is this fixed as part of the Kubeflow/Pipeline upgrade story?

[IronPan]

bootstrapper is now deprecated.