I just want to know what you guys think of this setup:
The goal is a HA mysql Server.
This is what it currently looks like:
| SWITCH |
| A |--| B |
Master to Master replication is done via direct gigabit connection
between machines A and B (cross cable).
Clients connect through the switch to the VIP.
mysql 5.0.19, auto_increment_increment=2, auto_increment_offset=1,
log-bin, log-slave-updates, server_id=1, master_host=B
keepalived 1.1.12, realserver: localhost, sorry_server: B, state MASTER
mysql 5.0.19, auto_increment_increment=2, auto_increment_offset=2,
log-bin, log-slave-updates, server_id=2, master_host=A
keepalived 1.1.12, realserver: A, sorry_server: localhost, state BACKUP
This would be some kind of "abuse" of keepalived. It is just intended to
take care of the VIP and put in the sorry_server on MySQL-failure. No
loadbalancing whatsoever. Machines are both big enough to handle the
amount of traffic.
On network failure on A, the VIP moves to B. Depending on the status of
the direct connection, either A or localhost is then used as MySQL-Server.
On MySQL-failure on A, machine B would be put in as sorry_server.
I think this could work out fine, what do you think? Any advice would be
appreciated. What pitfalls did I not think of?