Hi, I have a question on the HA proxy's high availability provided by the "peers" command,
I have two haproxy instances running - one as the master and another as the backup through keepalived, Both are configured to listen on a virtual ip and are servicing a couple of backend servers. The same configuration is running on both the instances. global daemon pidfile haproxy/haproxy.pid maxconn 40 defaults mode http balance roundrobin timeout connect 72000000 timeout client 72000000 timeout server 72000000 peers HA_peer peer LB-APPL 9.70.31.52:1981 peer SDN-VE-DSA 9.70.31.53:1984 listen 8080 bind 9.70.31.55:8080 mode http stick-table type ip size 20k peers HA_peer server web2 9.70.31.81:8000 check inter 2000 server web1 9.70.31.82:8000 check inter 2000 listen 23 bind 9.70.31.55:23 mode tcp stick-table type ip size 20k peers HA_peer server web2 9.70.31.81:23 check inter 2000 server web1 9.70.31.82:23 check inter 2000 I have an active HTTP session serviced through the haproxy's virtual ip and load balanced to one of the virtual servers. If the master haproxy instance goes down in the middle of the active HTTP session, can the backup haproxy instance that takes over the virtual ip ,know that the session was terminated in the middle ? Will it try to initiate a new session with the other backend server without the user having to intervene and start a new session at the front end ? I currently see that, when the master haproxy instance goes down, the active session is terminated. The backup haproxy instance takes over the virtual ip and only when the user tries to establish a new session, the backup haproxy connects the user to the backend server. Is this the expected behavior ? What is achieved through the "peers" and "stick-table" command here ? Please advise me on whether the configuration looks fine for this kind of testing. Thanks a lot for your response. Really appreciate it. Shweta
