OCP4: Haproxy reloads and long-lived connections deep dive
Issue
- OpenShift features the ingresscontroller system and router pods running a version of HAProxy coupled with frequent reloads. How does the reload system work, and how does this impact websocket traffic?
- I want to run a middleware application, database or other service on openshift with long-lived connections and I want to access them via route - what are some best practices to ensure this is successful?
- I am observing that long-lived connections are being terminated frequently via route - why?
- I am observing that long-lived connections are never closing on my router pods and I see CPU and Memory expanding on my router pods - how can I fix this?
Environment
- Red Hat OpenShift Container Platform (RHOCP) 4.x
- Ingresscontroller system (router-default or router-shard) in namespace openshift-ingress
- Traffic being managed by Routes
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.
