What are the best practices around replication & persistence (RDB snapshots and AOF)

[Andrew Feller]

Having used redis pooling largely and not redis replication, one of my colleagues asked about best practices around replication and persistence.  We're using redis-sentinel to handle master-slave promotion, so it's possible any given slave could become a master and vice versa.  The only commentary on the topic I've found talks about disabling / enabling persistence based upon the service's role, which is tricky in this scenario.  On top of all of that, we're writing RDB snapshots to an NFS mount and have to contend with whether all instances should read/write to the same file or not.  I'd be grateful for any insight anyone can offer!

Regards,
Andy