Hello everyone, according to the documentation, http://code.google.com/p/haproxy-docs/wiki/HTTPLogFormat the fourth element (Tr) indicates the service (processing) time, while Tt indicates the response time (i.e., service time + waiting time + transmission/network delay).
Suppose that the servers have a maxconn specified, e.g. server server1 <ip> maxconn 2 check inter 1000 rise 2 fall 2 slowstart 5s What happens when a response is ready (e.g., sent to HAProxy)? 1 - HAProxy computes Tr, sends the response to the client, computes Tt, and finally forwards another request to the server, which now has an idle slot, or 2 - It computes Tr, forwards another request to the server, then forwards the (first) response to the client and finally computes Tt 3 - It computes Tr, performs the two send operations in parallel, and computes Tt when the response forwarding is completed Thanks, Michele
