Exam AWS Certified Database - Specialty topic 1 question 190 discussion

Minimal downtime - operations first on replica and then promotion - Answer C.

A+D appear to be distractors, So there's an agreement that the DB instance class must be vertically up-scaled. The difference between B+C is that B applies it directlty to the Primary Instance and C to a Reader (replica) instance first. I believe that C is the wa to do it. https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/Aurora.Managing.Performance.html#Aurora.Managing.Performance.InstanceScaling

The thing here is that it is not stated whether we are looking at a Multi-AZ deployment. If that were to be the case, I understand that upgrading the type of the primary instance would cause very little downtime - just the time it would take to failover to the (upgraded) standby instance.

You can also modify DB instances in a DB cluster to accomplish tasks such as changing its DB instance class https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/Aurora.Modifying.html#Aurora.Modifying.Instance Hence, C can be apt for minimum downtime.

Answer from ChatGPT4 Based on the given scenario, the primary instance's CPU utilization is increasing, while the Aurora Replica CPU utilization remains steady. This indicates that the write workload has increased, whereas read queries are already well-distributed using the reader endpoint. To improve the database performance with minimal downtime, you can implement the following steps: 1. Scale the primary instance vertically by increasing its instance size: - Identify the appropriate Amazon Aurora DB instance size that can handle the increased workload and CPU utilization. - Modify the primary instance within the Aurora DB cluster to the new instance size. The failover process will cause a short downtime during the upgrade. The short downtime will smaller than the replica

The question reports that the CPU issue is on the primary instance and not on the RR.

C is wrong , You can't "Modify" to a large instance, You have to drop exist one then re-create a newer replica . so D is better than C

D is incorrect due to the possible higher RTO and RPO(higher Downtime). C is correct

C. There is minimal downtime with C, but NO downtime with D. The only issue is that D potentially up to date with the latest transactions (i.e. you used a backup).

D is correct