Murphy's Law influence on me
If there is an alarm monitoring, and already predict the seriousness of the event of the future, not to improve or do backup mechanism, then
sooner or later you will get a profound lesson.
(Murphy's Law)
Therefore, when you are not from redis clusters, while there has been an alarm, and then restored, but things are not likely to end there,
you should start to do backup solution, you can then cold standby for a cluster redis switching failure;
or you can upgrade the current cluster into the cluster redis redis3.x of a model (master-slave) of
In addition, the core underlying cache, storage and services, must be able to do more than write memory, places to live, clustered service,
short time to be able to fault the flexibility to quickly switch to ensure service availability. If a fault, and can only just wait, do anything, it is the software design of architect incompetence;