Once your replication topology is setup, it is important to consider the possible occurrences of failures in order to maintain the high availability level of your system. Several failures independently from their cause (network connection issue, hard drive crash, cosmic lightning, etc.) can stop the replication process by making the master no longer accessible by its slaves.
In this type of situation, it is desirable to promote one of the slaves to master while the problem with the old master is being solve as it can take some considerable time. It will be even better to have an application to monitor the replicate topology and automatically perform failover, minimizing downtime and keeping replication running.