How does DocumentDB failover happen if the the reader instance and writer instance have different instance types?

Question

I have a DocumentDB cluster with exactly one replica instance ( db.r5.16xlarge ) and a primary instance ( db.r5.24xlarge ). If something happens to the primary instance and a failover occurs, the replica instance will act as the new primary instance. Once the previous primary instance becomes ready, will it again failover back to my initial db.r5.24xlarge primary instance?

Answer 1

No, it will not failover again to promote the previous primary instance. You'll have to failover manually to promote the bigger instance.

Better control can be achieved with 3 instance using failover tiers. Each replica instance is associated with a failover tier (0–15), where a lower tier means the instance has higher priority to be chosen for election.

For example, a cluster with two db.r5.24xlarge set on tier-1 and one db.r5.16xlarge set on tier-2. If a primary db.r5.24xlarge becomes unavailable, then the other db.r5.24xlarge will be promoted because it has a higher priority.

Reference.