Hi haproxy community,
I ask here from advices for a very specific haproxy configuration.
For the context we are in a niche business were have to handle million of request/s but without the need answering them (from backend perspective) to all of them ; meaning it’s perfectly ok not to treat them all.
But there is a big but we absolutely need to answer fast (sub 200ms) at least 204 to all of them (from client perspective).
So what I try to achieve is to mask every error possible returned by backend. and also try to mask backend slowness.
Also with this volume we need to tune haproxy to perform.
I struggle a bit with all timer and config.
We have a config that kinda work but still producing some client facing error (50x we want to avoid).
I have a mix of errorfile XXX /etc/haproxy/204.http interception, small timeout, but still we got errors transmitted from backend to client.
TLDR, If possible I want something like replying 204 to any error possible , and ideally to handle a fixed amount of request per second from a backend section (so hundred of backend servers).
I can provide clarifications if you want.