Using_With_Clustering_Software

Using with clustering software

If you use a virtual IP address for a master, you might already use clustering software like Pacemaker. If you are familiar with existing clustering tools, you may want to use them to manage virtual ip address, rather than doing everything in MHA. MHA can be used just for (non-interactive) failover, so you can use clustering software and MHA together for master failover.

Clustering software manages a virtual IP address on the master and does node fencing. MHA focuses on promoting a new master and fixing slave consistency.

Here is a brief example for configuring Pacemaker (Heartbeat v1 mode).

# /etc/ha.d/haresources on host2
host2  failover_start  IPaddr::192.168.0.3

failover_start script example

start)
masterha_master_switch --master_state=dead --interactive=0 --wait_on_failover_error=0 --dead_master_host=host1 --new_master_host=host2
exit

stop)
# do nothing

Application configuration file:

[server1]
hostname=host1
candidate_master=1

[server2]
hostname=host2
candidate_master=1

[server3]
hostname=host3
no_master=1

Since data files are not shared, data recources are not needed to be managed by clustering software or DRBD. For this purpose, clustering software just invokes masterha_master_switch script and takes over virtual IP address. You can do the same thing by handmade scripts etc.