ROUTER_SLOWLORIS_TIMEOUT is currently a read only option in OCP4 router configuration, see https://docs.openshift.com/container-platform/4.11/networking/routes/route-configuration.html

It was possible tochange this value in OCP3.

Customer Requirement:

At peak times, we have the problem that some applications get connection reset errors. Changing that HAProxy parameter "timeout http-request" does correct the issue.

Changing the value is recommended by upstream haproxy in case of HTTP 408 errors, which has been seem in some customer setup.
see https://www.haproxy.com/blog/haproxy-and-http-errors-408-in-chrome/

The Red Hat Product team has already introduced a few tuning parameters, but this parameter is not exposed like many other parameters available in the route image and no technical and documented reasons are given.

Affected Component

Ingress Operator, router pods

Relation to other RFE

We want to allow in https://issues.redhat.com/browse/RFE-3347 L7 load balancers.
Those load balancers will probably have a kind of pooling behaviour similar to

https://www.nginx.com/blog/load-balancing-with-nginx-plus-part-2/

if this pooling behaviour is implemeted at the HTTP level, then HAProxy will reject each connection in the pool which is unused after 10s with a 408 error.

Knowns workaround

Currently the only know workaround to get this OCP3 option working on OCP4 is to create an override for the Cluster Version Operator, which makes the whole cluster unsupported.