1. Proposed title of this feature request
OCP 4.13+: Full support of http/2 and websockets

2. What is the nature and description of the request?
Goal:

• Full HTTP/2 without websocket limitations for http/1.1 clients
• HAProxy implemented in v2.4.0: https://github.com/haproxy/haproxy/issues/162

Current limitation[1]:

• "Using WebSockets with a re-encrypt route and with HTTP/2 enabled on an Ingress Controller requires WebSocket support over HTTP/2."
• "WebSockets over HTTP/2 is a feature of HAProxy 2.4, which is unsupported in OpenShift Container Platform at this time."

Previously we had a Bug[2] open to get "full support" with haproxy 2.2, which was closed with "Won't Do".
"Marking as deferred as this would be fixed with HAProxy 2.4."

OCP 4.13+ was initially planned with haproxy 2.6, which should have included the feature. Due to a bug found during haproxy 2.6 testing it was reverted back to haproxy 2.2 (for OCP 4.13+). References[3][4][5].

3. Why does the customer need this? (List the business requirements here)
Customer needs this feature to onboard workload from customers. Without this feature, projects have to look for possible solutions outside of Openshift.

4. List any affected packages or components.
HTTP/2 Ingress connectivity

References:
1) Enabling HTTP/2 Ingress connectivity

• https://docs.openshift.com/container-platform/4.12/networking/ingress-operator.html#nw-http2-haproxy_configuring-ingress

2) With HTTP2 enabled, routes using reencrypt can't use websockets

• https://issues.redhat.com/browse/OCPBUGS-173

3) OCPBUGS-10003: Revert "NE-1115: Update haproxy container builds to use haproxy 2.6

• https://github.com/openshift/router/pull/458

4) OCPBUGS-10003: Revert "bump RHEL8 egress-dns-proxy image to haproxy26"

• https://github.com/openshift/images/pull/134

5) [reliability]one router-default pod's haproxy cpu usage suddenly increased

• https://issues.redhat.com/browse/OCPBUGS-10003