Hi, I have a few questions around setting up a customized haproxy farm. I am combining all my questions in single email.
1) Map peering Is it possible to peer maps across haproxy running on multiple machines? Or store key, value pairs inside stick table (which supports peering). 2) Backend of backends I am using haproxy to load balance between multiple backends (A/B testing). I am currently creating a new frontend for each backend in the A/B/C test which listens on unix socket. Based on previous email threads, I believe that listening on TCP socket gives better performance than listening on unix sockets. So frontend listening on 443 forwards request to frontend listening on unix socket which then forwards the request to actual backend servers. Would listening on localhost tcp ports (instead of unix sockets) be the recommended approach? I am managing multiple A/B tests using same haproxy farm. So, do I have to ensure that test X gets the same port across haproxy soft restarts? I wouldn't want test 34 traffic to forward to test 35 because test 34 port got assigned to test 35 when config file was rewritten using chef. 3) Call URL before forwarding Is there any way to call a specific URL to validate the request and then forward the request based on response from the previous call? I am using this to see if user is logged in. If the user is not logged in, I want to redirect the user to login screen. Does another reverse proxy engine like Nginx have such feature? Regards, Vivek
