Doing another startup which used the radish, now closely was the question of building a failover database cluster.
As you know, radish supports only master-slave replication. It is necessary that at falling of the master some of Slavov would take a wizard function. For this purpose I found the following solution:
github.com/fictorial/redis-cluster-monitor. In the fall of the master, this monetarily chooses a new master and sends a signal synchronization with the other servers. Task — to notify the frontends that the master server has changed. Going to finish monitoringu that she had sent notices and frontends. Address of the master server will probably be stored locally in a file.
How many correct this approach? Anyone had experience building such systems, what is usually done, what are the pitfalls?
