When used in combination with MetalLB, users want a way to choose what pods run on a given node. See https://issues.redhat.com/browse/OCPBUGS-56173?focusedId=27188511&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-27188511 where a customer chooses where to run metallb speakers from. After moving frr-k8s to be deployed by CNO, this is not possible anymore. So, a user can use the node selector of MetalLB to choose where to run the speakers, but CNO will deploy frr-k8s pods on all the nodes.    

    Always

    1.
    2.
    3.
    

    Having a way not to schedule FRR-K8s pods on some nodes. This can be done with a full fledged node selector or a "don'tschedulefrrk8shere" sort of well know label. I think it's more than reasonable not to use the same node selector of metallb, while documenting that the set of nodes running the speakers and frrk8s pods must be the same.

    Preliminary work was done as part of https://issues.redhat.com/browse/OCPBUGS-56173, to provide OVN-Kubernetes a way to understand if an frrk8s pod is running or not on a node by looking at the presence of the frrnodestate instance corresponding to that node.