Description
We have a bunch of different traffic flowing through our application, all of them with different characteristics (latency tolerance, retry resilience, impact on user experience, criticality to proper application functioning, etc.), all of them are identifiable through a variety of different matchings on the request itself.
We have a kubernetes cluster, where currently an nginx ingress controller distributes requests across a variety of api servers. Upstream to that we have another nginx server (frontend) doing the matching, and sending traffic into the kubernetes cluster. We use kubernetes for this, as the volume of traffic ebbs and floods over the day, and auto-scaling on web worker usage and queue length has proven to be an effective method of handling with this.
We want to switch over to using the haproxy ingress controller to take advantage of request prioritisation, as this fits our usecase for managing the above mentioned different traffic patterns. For simplification, the plan was to have the upstream nginx ingress "mark" the requests as such before sending them along, so that a simple haproxy priority class matcher could assign it priority in the queue.
However, it appears that the haproxy ingress controller has not implemented the priority mechanism, as I couldn't find any mention of it in this repository, nor in documentation.