Hi, How do you need to design AWS RDS/Aurora deployment to achieve RPO of zero?
That’s a good question as I’m not sure you can achieve zero RPO with RDS and Aurora because any multi-az or multi-region deployment will almost always require an async replication component….and that introduces some lag for the replica…maybe only a few seconds but it’s still there.
Supposedly, Aurora is designed to provide zero RPO within a single region but AWS is very careful to use near-zero in most documentation.
There is a specific failure mode for RDS that can maintain zero RPO…if an instance fails, RDS will spin up a new instance and just attach to the EBS volume…with no (theoretical) data loss. However, we don’t get to choose our failure modes.
There are new features being introduced every day and maybe someone else knows of a way to do this. If so, comment below!