Hands-On Software Engineering with Golang

Ensuring that masters are highly available

In a system built using the master/worker model, losing one or more worker nodes due to crashes or network partitions is not a big issue. The master can detect this and work around the problem by re-distributing the workload to the remaining workers.

A crucial piece of advice when designing distributed systems is to make sure that your system does not contain Single Points of Failure (SPoFs).

On the other hand, the loss of the master node will most certainly take the entire system offline! Fortunately, there are a few different approaches at our disposal for making sure that master nodes are highly available, which we'll cover next.