Hi, I have a deployment with two HAProxies in a peers configuration. A normal flow is that a client issues two subsequent calls (First response from backend servers sets a key on the sticky table and the second requests uses that key)
Because the HW LB balancing between both HAProxies works in round robin, it's highly likely that first call will go to one HAProxy and second call will go to the others. The problem that I'm facing is that the second request seems not to find the key on the stickiness table and thus the requests is sent to any backend server. Worth to mention that this only happens the first time a pair of this calls is made. Subsequent calls already have the key on the stickiness table and works fine. The problem is exacerbated by the fact that I'm adding and removing backend servers to HAProxies oftenly so everytime I add a server the first client that ends in that server will have is second call fail. As suggested in this forum, I first add the backend server as disabled in both HAProxies, then I enable it in one and finally enable it in the other. Any ideas? peers section: peers cluster peer haproxy--i-46279671 172.18.80.136:1024 peer haproxy--i-b8e68a8c 172.18.81.126:1024 backend section backend farm option httpchk GET /health HTTP/1.0 option forwardfor balance roundrobin stick store-response hdr(X-Session) stick-table type string len 24 size 200k peers cluster stick match url_param(X-Session) server vm--i-37a49303 172.18.80.106 check port 80 inter 30s fastinter 5s rise 2 fall 3 server vm--i-43279674 172.18.80.60 check port 80 inter 30s fastinter 5s rise 2 fall 3 Thanks! Jose *J+*
