NTP out of sync issue with 3 server XtraDB cluster

We’ve discovered that the time on each of 3 Ubuntu (AWS) servers running in a XtraDB cluster are now badly out-of-sync - by as much as several minutes between them. Is it at all possible to reset the time while XtraDB is running - or do the DB instances need to be systematically stopped one by one before resetting the time and restarting, and the cluster members restarted in some sort of order?
Are there some best practices to check and ensure that the time is appropriately in sync between such servers - and if they drift too far, what to do to bring them back in sync?