Hey there, let's break down your questions about Aurora's high availability setup clearly:

1. Does Aurora's MultiAZ configuration create a DB instance in every AZ within a region? #

No, it doesn't. Aurora's default MultiAZ setup automatically deploys one primary DB instance and one synchronous standby replica in a different AZ within the same region. This standby is maintained by RDS for failover purposes, but it won't automatically spin up instances in every available AZ in the region—you'd have to manually add additional read replicas to other AZs if needed.

2. If I currently have three AWS::RDS::DBInstance resources each in a separate AZ, do I need to keep all three for high availability? #

From a pure high availability standpoint, no. The core Aurora MultiAZ setup (primary + cross-AZ standby) already provides automatic failover capability if the primary instance or its AZ goes down. Your three instances are likely adding value for read throughput and performance scaling (offloading read queries to replicas), but they aren't required to meet basic high availability requirements.

3. Can a single instance meet high availability needs? #

It depends on what you mean by "single instance":

• If you're referring to an Aurora cluster configured with MultiAZ (which actually includes a primary instance and a hidden synchronous standby in another AZ), then yes—this setup meets high availability standards, with RDS handling automatic failover typically within 30 seconds.
• If you're talking about a single, non-MultiAZ instance (only in one AZ with no standby), then no. This setup has no failover mechanism, so if the instance or AZ fails, your database will be unavailable until it's recovered.

4. For high availability alone, do I need to deploy three instances across three AZs? #

Nope. Aurora's native MultiAZ architecture (primary + one cross-AZ standby) already delivers the 99.99% availability SLA that AWS guarantees for MultiAZ deployments. Adding extra instances across more AZs is beneficial for scaling read performance, distributing read traffic geographically within the region, or adding redundancy beyond the basic failover setup—but it's not a requirement for core high availability.

内容的提问来源于stack exchange，提问作者Joey Yi Zhao