My client claims that Aurora Serverless V2 guarantees an RPO of less than five minutes in the event of a disaster. Is it correct? I cannot find a clear reference except for Global Aurora (which does not correspond to its choice).
If yes, is this correct regardless of the configuration of backups/snapshots and Multi AZ deployment?
More precisely, does the choice between Single AZ and Multi AZ influence the RPO for Aurora Serverless V2?
The Recovery Point Objective for an Aurora DB depends on the "disaster" you assume:
Instance failures and AZ-wide outages
AWS Aurora stores data in cluster volumes. These cluster volumes are replicated over 3 Availability Zones at the storage level. (see AWS docs). The replication happens via sync writes, so your data is stored in 3 AZs when committed to Aurora.
In case of a failure in the DB instance, an error in a single disk, or an outage on AZ level, the Recovery Point Objective is close to 0.
Region-wide outages
When Automated Backups are enabled for the database cluster (which should be the default), Aurora continuously backs up your data to S3. Following the AWS docs, this happens at least every 5 minutes (see source in AWS docs).
In case of a Region-wide failure of the volume cluster, the Recovery Point Objective is defined by the max. delay of Backup synchronization, which is 5 minutes.
Multi AZ Deployments
Regarding your follow-up question with Multi AZ Deployments: In this setup you typically deploy multiple Aurora instances in different AZs in the same Region. This provides a hot standby instance for the case that the primary instance fails, which reduces the Recovery Time Objective. All instances work on the same volume cluster as storage. So this has no influence on the Recovery Point Objective.
Hope that helps - happy to discuss that further.