Why only 2 KeepAliveD?
Any reason it is only possible to deploy 2 Keepalived instances in Clustercontrol? What if I want a VIP for 3 HAproxy nodes...?
ClusterControl implements Keepalived's virtual IP address with master-backup approach, which has the least requirement of two HAproxy instances. Only one HAproxy instance holds the VIP in a time. If you had 3 HAProxy instances, and the first one went down, the VIP would be failed over to haproxy #2 (the backup instance). Once haproxy #1 got back up, the VIP would be failed back to haproxy #1 (the master instance). Haproxy #3 won't be used unless #1 and #2 go down at a same time.
In most cases, having two HAproxy instances is suffice in production environment (provided the HAproxy instances live in separate physical hosts).
To answer your question, you can just install Keepalived manually on HAproxy #3 and configure /etc/keepalived/keepalived.conf with similar configuration as the rest. Ensure on #3, the state is BACKUP, holds the same VIP and the unicast peer list is updated accordingly.
Please sign in to leave a comment.