Re: Failover for SQL nodes?

[Harrison Fisk]

Hi,

There are a couple of options here. Basically since you can query against any of the MySQL nodes, it makes failover quite easy.

1. Setup a single mysqld for each application server. You can run them in a 1-to-1 setup, normally locally on the same machine. That way all of the data retrieve is local, with only remote data being queried internally.

2. The JDBC driver can handle this automatically. The most recent 3.1 releases can handle all of the failover for you. One could easily create wrappers for other APIs to do it as well (it will most likely be added to our standard APIs at some point).

3. DNS. It would be fairly easy to setup round robin DNS and remove an entry if the server goes down.

4. Other software such as Linux Virtual Server or Ultra Monkey. These can handle failure situations and take-over setups pretty easily.

Of course there are other ways you can do it, such as hardware solutions (Big-IP or similar). Hope that gives you some ideas how to do it.

Harrison Fisk, Trainer and Consultant
MySQL AB, www.mysql.com