We are attempting to use the appsession config option to ensure that requests with the same session id in the url query parameters go to the same backend. We are also using the "option redispatch" setting to ensure that sessions get sent to a new backend when an existing backend goes down. It seems that enabling "option redispatch" is eliminating the stickness that appsession was having even when the backends are not going down. Is this a known issue? All of our http requests are also using keep-alive.
Here is the relevant portion of our haproxy config: defaults retries 3 option redispatch mode http timeout connect 5000 timeout client 50000 timeout server 50000 timeout check 10000 frontend ws_in bind 127.0.0.1:8080 option httplog default_backend ws backend ws option httplog balance roundrobin appsession session_id len 52 timeout 3h request-learn mode query-string server ws_1 localhost.org.:8030 server ws_2 localhost.org.:8031 ------ Thank you, Kris