Exploring adding support guidance around Reference Architectures split over an odd number of Availability Zones as a form of Disaster Recovery
Recently we've been hearing from customers the desire to use Availability Zones (AZ) in the same region as a form of Disaster Recovery (DR).
This is a complicated area but one such solution presented looked quite viable where the customer was looking to effectively take a HA Reference Architecture and split it across 3 nodes. By doing it in this way they're able to maintain the correct number of components for a Quorum and not require a full second environment in the case of a disaster.
That said with this approach the following caveats apply:
- That nodes are able to communicate over AZs
- AZs should be in the same location (can be different buildings, etc…)
- That network speeds across AZs are excellent
Task is to explore if this is correct and add guidance around this in the docs.
Edited by Grant Young