Description of problem:

From OCP4.19 frrk8s pods are deployed in a openshift-frr-k8s namespace instead of metallb-system. Earlier speaker and frrk8s pod co-exist on a node. With the new change, frrk8s pods run in all nodes even without a speaker pod. 

Because of this, when a speaker is removed from a node, FrrConfiguration still has the previous BGPPeer and other configuration still in place. 

Version-Release number of selected component (if applicable): 4.19

How reproducible: 100%

• Steps to Reproduce:
• Create MetalLB without nodeselector
• Create BGPPeer with peer ip1
• Notice FrrConfiguration is updated for all nodes with bgp peer ip1
• Delete MetalLB
• Create MetalLB with NodeSelector for worker nodes
• Delete BGPPeer
• Notice FrrConfiguration is updated only for worker nodes with empty spec
• Notice FrrConfiguration for master nodes still has previous peer ip1
• Create BGPPeer with peer ip2
• Notice FrrConfiguration is updated for worker nodes with peer ip2
• Notice FrrConfiguration for master nodes still has previous peer ip1

Actual results: Stale FrrConfiguration exists

Expected results: Removing speaker pod from a node should trigger FrrConfiguration reset 

Additional info: