Re: Failover without Cluster

[James Day]

For the future there are plans to make this more automatic. For now, the general approaches used involve monitoring scripts which check to see that the server is alive and switch master if it isn't. If you've many nodes and different applications a service IP approach and IP address switch may be most effective. Alternatively, you may have a configuration file whcih can be swapped to move all services to the new master. Remember to change the new master to read-write if you have it set to read only.

A service IP is an IP identifying a particular service which can be moved around to different computers as needed to switch which one is doing the job. Particularly helpful if many different applications are involved and it's a lot of work to make the switch in in many configuration places.