Today ovn-org/ovn-kubernetes executes the following CI jobs on each new PR:

https://github.com/ovn-org/ovn-kubernetes/blob/956f8fa6b86858c15659c1e4f22b06ab9ebae21a/.github/workflows/test.yml#L409-L439

These use an ovn-kubernetes image built from the PR code and using fedora OVS/OVN rpms:

https://github.com/ovn-org/ovn-kubernetes/blob/956f8fa6b86858c15659c1e4f22b06ab9ebae21a/dist/images/Dockerfile.fedora#L18

At the same time, on each patch submission / PR / push, ovn-org/ovn executes a subset of the ovn-kubernetes CI jobs:

https://github.com/ovn-org/ovn/blob/main/.github/workflows/ovn-kubernetes.yml#L79-L85

These instead use a custom Dockerfile that builds an ovn-kubernetes image using the patched OVN version:

https://github.com/ovn-org/ovn/blob/main/.ci/ovn-kubernetes/Dockerfile

This duplication of efforts is not good and also means the ovn-org/ovn CI will always be behind ovn-kubernetes CI in terms of infrastructure changes (go versions, k8s versions, etc.) but also when it comes to identifying tests that occasionally fail (flaky tests).

The goal of this epic is to track the work needed for consolidating the two CI efforts.