if you're using physical replication/DR-style replication,

datastore repair

won't do anything, and the solution is in other configuration