Skip to content

helm upgrade from manifest-installed v1.2.x wedges: Fail-policy preset webhooks applied before the new manager serves them #835

Description

@CatherineSue

Problem

Upgrading a manifest-installed OME v1.2.1 to main via helm upgrade --install --take-ownership applies the new servingruntime-preset.ome.io MutatingWebhookConfiguration (failurePolicy: Fail, path /mutate-ome-io-v1beta1-servingruntime-preset) before the ome-controller-manager Deployment rolls to an image that serves that path. The chart's own default-runtime ClusterServingRuntime then trips the webhook against the still-running old manager:

Internal error occurred: failed calling webhook "clusterservingruntime.ome-webhook-server.preset": failed to call webhook: the server could not find the requested resource

The release fails mid-apply and — worse — ServingRuntime/ClusterServingRuntime admission is broken cluster-wide until the manager is manually rolled forward (we had to kubectl set image by hand, then re-run helm).

Also hit on the same path: the retired ISVC conversion webhook leaves spec.conversion + cert-manager.io/inject-ca-from on the live inferenceservices.ome.io CRD, which makes a server-side apply of main's conversion-less CRD invalid (spec.conversion.strategy: Required value) until the stale conversion stanza and annotation are removed by hand.

Suggested fix

Ship the preset webhooks with a helm pre/post hook ordering (or failurePolicy: Ignore until the manager is rolled), and document the v1.2.x → main CRD migration (conversion stanza removal) in upgrade notes.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions