backup – Highly available mysql in kubernetes

By following the documentation “Run a Replicated Stateful Application” I created 1 master mysql and 2 slaves, when I try to kill the master with kubectl exec mysql-0 -c mysql -- mv /usr/bin/mysql /usr/bin/mysql.off it is not possible to write because mysql-0.mysql is down, only read is available from the replicas, this makes the website not funcional, is there a bug or it is supposed to work like this? From my experience I know on this topology when master is down a slave becomes master. What is the benefit of just reading without writting?