Description of problem:

When ovnkube-control-plane starts up if two egressIP CRs have the same egressip-mark mark annotation, the control-plane pod goes into crashloop backoff.

2025-xx-xxTxx:xx:xxZ E0820 01:17:44.002178       1 factory.go:1301] Failed (will retry) while processing existing *v1.EgressIP items: failed to reserve mark for EgressIP <name>: failed to reserve mark: id 0 is already reserved by another resource
--- many repeats ^ --
ovnkube.go:593] Stopping ovnkube

oc get egressip -A -o yaml | grep egressip-mark | cut -d: -f2 | sort  | uniq -c
      2  "50000"
      1  "50001"
--snip--

Version-Release number of selected component (if applicable):

Steps to Reproduce:

1. Make two egressIP CRs with the same egressip-mark annotation value.

Actual results:

ovnkube-control-plane crashes.

Expected results:

ovnkube-control-plane starts up.

Additional info:

We are not sure how these duplicate annotations occur; however, they prevent the startup of ovnk cluster-wide.

resolution.

Affected Platforms: Customer Cluster

Workaround:

This removes the annotation from all egressips, allowing the kubernetes-ovn control-plan pods to start up sync and then they will reallocate the mark without duplication. If the EgressIP is managed via gitops/argocd the customer may need to pause the applications  to prevent the re-application of the duplicate mark value.

oc annotate egressip -A --all k8s.ovn.org/egressip-mark-