OCP4: LLDP configuration enabled via nmstate/nncp breaks OpenShift node after restart
Issue
- An existing bond
bond0(or other primary bond for node networking) is updated to includelldp=1as an optional argument via NodeNetworkConfigurationPolicy (NNCP) update, or manually on the host - The object updates successfully, Nodes remain in READY state
- After node Reboot, it does not return to READY. Networking is down on the host, and/or the bond is removed/lost/mangled/missing.
- A new interface:
ovs-if-phys0.nmconnectionexists in the directory/etc/NetworkManager/system-connections/which did not exist previously and may include bits of the previousbond0interface used as the primary interface of the host. - Networking is down/impacted.
Environment
- Red Hat OpenShift Container Platform (RHOCP) 4.x
- OVN-Kubernetes with Automatic OVS configuration script initialization (automatically configured br-ex interface via ovs-configuration.service)
- Bonding with LLDP enabled
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.