The document describes the MHA failover process in 5 phases:
1) Configuration check phase verifies the state of nodes
2) Dead master shutdown phase disables the dead master
3) Master recovery phase determines the new master from candidates and applies binary logs from the dead and latest slaves
4) Slaves recovery phase generates and applies relay log diffs to synchronize slaves with the new master
5) New master cleanup phase sets the new master read-only attribute to false after failover is completed