SAA-C03 ยท High availability and disaster recovery
Highly available relational databases on AWS
RDS Multi-AZ deployments and failover, Aurora replicas across AZs and Aurora Global Database for cross-Region recovery.
Key points
- Multi-AZ DB instance: primary + synchronous standby in another AZ, no reads.
- Failover: DNS flips in 60-120 s; test with reboot with failover.
- Multi-AZ DB cluster: writer + two readable standbys, three AZs, semisynchronous.
- Read replicas: scale reads, asynchronous.
- Aurora: shared volume, up to 15 Aurora Replicas, promoted on failover.
- Aurora Global Database: up to 10 read-only secondary Regions, lag under a second, promotable.
Source: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Concepts.MultiAZSingleStandby.html
Practice questions
How does an RDS Multi-AZ DB instance deployment keep its standby up to date?
- Asynchronous replication, like a read replica
- Synchronous replication to a standby in another AZ
- Hourly snapshots restored onto the standby
- Transaction logs shipped to S3 every five minutes
Show answer
Answer: B. Synchronous replication to a standby in another AZ
The primary replicates synchronously to the standby in another AZ, so committed data is already on the standby when a failover happens. Asynchronous replication is how read replicas work; snapshots and log shipping to S3 are backup mechanisms, not the standby's replication.
Which events cause an RDS Multi-AZ DB instance to fail over to its standby automatically?
- The primary host becomes unhealthy
- Network connectivity to the primary is lost
- A single query runs for longer than a minute
- The primary's storage volume fails
- The application sends SQL with a syntax error
Show answer
Answer: A. The primary host becomes unhealthy, B. Network connectivity to the primary is lost, D. The primary's storage volume fails
RDS fails over when the primary host is unhealthy, when it becomes unreachable through loss of network connectivity, and when its storage volume fails (also during OS patching and on a reboot with failover). A slow query or a SQL error in the application is a workload problem; the primary is still healthy, so no failover happens.
After an RDS Multi-AZ failover, the application must be reconfigured with a new database endpoint name that points to the former standby.
Show answer
Answer: False
False. RDS updates the DNS record behind the same endpoint to point at the new primary. The application only has to re-establish its connections, and should not cache DNS lookups for too long.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
