I have a frontend in mode tcp that I use to switch inbound traffic between a syslog backend(raw tcp) and an http backend.
I've been using this setup for over a year with little issue, but I think I've been missing a case that is now more prevalent with more traffic. Given the frontend config: https://gist.github.com/davidbirdsong/b39a98567bc33930004f3e87ba5af428 New connections match on the api acl, but it seems(we're still trying to isolate) that subsequent requests on the same TCP connection do not have the default "HTTP" acl set to true. Is there a weakness in this logic that can be shored up?
