1. Proposed title of this feature request
Make Service External IP highly available again.

2. What is the nature and description of the request?
Besides router and NodePort service external IP is a common way of routing traffic into the cluster for non cloud environments. This is very useful when the traffic is not http and has no SNI.
To use a service external IP it is required that the allocated IP address is configured on a node. To make it highly available this IP address needs to be automatically and quickly reallocated to a different node when the first node is not available. In 3.x this was achieved by using a failover IP and the image provided for this purpose. This mechanism is currently not supported with OCP 4,. cf: RFE-451

3. Why does the customer need this? (List the business requirements here)
Binding the address directly to a single node makes the service subject to outages:

• when the server has an issue
• when there is a planned maintenance
• every time an OCP upgrade is rolled out and the server is rebooted. Mind that this can take between 20 and 30 minutes with big physical machines. This is a significant hurdle to OCP 4 aim of making upgrades easy and transparent.

Reconfiguring the service external IP to a different machine is then a manual task with all the usual drawbacks. Mind that customers may have hundreds of external service IPs that they will allocate to different nodes for spreading the load.

4. List any affected packages or components.
Possibly old/unsupported IP failover image

Additional information:

• This RFE is for OCP 4.x
• It is about highly availability not load balancing
• It would be nice to have an easy way of getting the configuration applied, there is no specific requirement whether this should be done through CLI, machine config or other custom resources.