I can't find in documentation any clear information about haproxy behavior on failed connection in tcp mode.
I have configured 2 backends, balance roundrobin option and health check for both backends.
And when one of the backend fall down, but health check haven't yet marked it as down it is possible that haproxy will try to send new connection to it. And will receive timeout on connection attempt. And what will haproxy do in this situation?
Will it try to connect retries times to this server again? Or will it try another one?
As i understand from documentation option redispatch works only for http mode. So haproxy will not try to connect to other backend? If it so, why option redispatch not works for tcp?