Taking a snapshot in a Multi-AZ deployment will not affect the availability of the database. With a Multi-AZ deployment, the snapshot is taken from the secondary instance while the master is still available to take queries and writes. We should consider that, when many writes are happening on the master instance during snapshot creation, those need to be replicated after the snapshot completes to the secondary, thus we might see a spike in latency in our CloudWatch monitoring due to the fact that the secondary instance is being replicated after the snapshot completes.

In Amazon Aurora, snapshots are even more transparent. Aurora instances are stateless, so a snapshot of the backend cluster storage is taken during an RDS snapshot operation. Aurora snapshots are thus referred to as cluster database snapshots.