Currently, when we make an ovn change, we run the ovn-kubernetes testsuite against OVN using this change. We use the master version of ovn-k.

The problem with master is that it means that we are testing against an inconsistent version of ovn-k. A bug may have been introduced to ovn-k that causes the testsuite to fail.

Instead, we should pick a specific "stable" version of ovn-k and run against that for a period of time until we decide it is time to upgrade to a new version. This will make debugging test failures more consistent since we will not have a moving target for ovn-k.

ovn-k does not create tagged releases, so we cannot rely on the ovn-k project to determine what is a stable version. Instead, we will need to pick an arbitrary hash, perhaps one that is used in specific releases of Red Hat OpenShift.