High Availability for Redis Cluster

[Alexander Cook]

Hi, 

I am interested in learning more about the future of Redis Cluster with respect to High Availability. There is some degree of it available, but generally Redis Sentinel is currently the only option for strong guarantees of high availability. 

For example, with a 3 node, 3 master, 6 slave setup (starting one master per host, then slaves replicating it on the two other hosts). We start out with a setup that is tolerant to the failure of a single host. 

Host 1 fails, then it's slave on Host 2 or 3 takes over as the master for that shard. 

If Host 1 resumes, and its Redis instances join the cluster again, there is nothing that automatically returns the master to Host 1 (right?). So in this event, although we have 3 hosts, unless a manual failover is done, aren't we now intolerant to any failures? Are there any plans/ways to address this? An option would be to write a script that manually checks to make sure things are distributed, then failover if necessary. 

I'm new to the group, so getting pointed to discussions for the future of Redis Cluster would be great. Thanks!