1. Proposed title of this feature request
Support for Multiple (2 or above) API and Ingress VIPs in install-config.yaml for Multi-AZ OpenShift on OpenStack Deployments
2. What is the nature and description of the request?
Currently, the OpenShift installer only allows a maximum of two entries (one IPv4 and one IPv6) for both apiVIPs and ingressVIPs in the install-config.yaml file when deploying on OpenStack.
We request support for specifying multiple VIPs (e.g., 2 IPs × 3 AZs = 6 VIPs for dual-stack across three availability zones) in both apiVIPs and ingressVIPs.
This would enable customers to assign a dedicated VIP to each Octavia load balancer deployed per AZ, and then use DNS round-robin (via Designate) to distribute traffic, thereby eliminating the single point of failure (SPOF) at the load balancer level in multi-AZ architectures.
Octavia uses VRRP for Amphora redundancy, and it is assumed that Amphora instances are connected to the same L2 network even if they are in different availability zones (AZs). However, in the customer’s environment, L2 networks cannot span across AZs, so it seems that Octavia’s multi-AZ feature would not be effective.
3. Why does the customer need this? (List the business requirements here)
- To provide a highly available, multi-AZ OpenShift as a Service (KaaS) platform to internal users.
- To eliminate load balancer SPOF by deploying an Octavia LB in each AZ and distributing VIPs via DNS round-robin.
- To fully leverage OpenStack's multi-AZ capabilities for both control plane and data plane, improving resilience against AZ-level failures.
4. List any affected packages or components.
- openshift-installer (specifically, the OpenStack platform integration)
- OpenShift documentation (installation and networking guides)