SAA-C03 ยท High-performing databases and caching
Read replicas: scaling reads and cross-Region copies
Offloading read-heavy traffic with RDS and Aurora read replicas, asynchronous replication lag, cross-Region replicas and promoting a replica.
Key points
- RDS read replica: read-only copy built from a snapshot, updated by asynchronous replication.
- Offloads reporting and read-heavy traffic; may lag and return stale data.
- Another AZ or Region (no cross-Region for SQL Server); promotable to a writable standalone instance for DR.
- No automatic adding or removing; billed as a normal DB instance.
- Aurora: up to 15 Aurora Replicas on the shared volume, lag under 100 ms, failover targets.
Source: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_ReadRepl.html
Practice questions
How does Amazon RDS keep a DB instance read replica up to date with its primary?
- Asynchronous replication using the engine's built-in mechanism
- Synchronous replication that commits on both before acknowledging
- A nightly snapshot restored onto the replica instance
- Shared EBS volume mounted by both primary and replica
Show answer
Answer: A. Asynchronous replication using the engine's built-in mechanism
After the initial snapshot, RDS uses the engine's asynchronous replication to send changes to the replica. Synchronous replication is what a Multi-AZ standby uses, and that standby cannot serve reads.
Which statements about Amazon RDS DB instance read replicas are correct?
- Writes sent to a replica are synced back to the primary
- A read replica is billed as a standard DB instance
- A read replica can be placed in a different AWS Region
- RDS adds replicas automatically when read traffic grows
- A read replica can be promoted to a standalone DB instance
Show answer
Answer: B. A read replica is billed as a standard DB instance, C. A read replica can be placed in a different AWS Region, E. A read replica can be promoted to a standalone DB instance
Replicas can sit in another Region, can be promoted to a standalone instance, and are billed like any DB instance. RDS does not auto scale the number of replicas, and a replica accepts only read-only connections; writes go to the primary.
A query on an RDS read replica always returns exactly the same data as the primary at that moment.
Show answer
Answer: False
False. Replication to RDS read replicas is asynchronous, so the replica can lag and return data that is slightly out of date, especially during heavy write periods. Apps that must read their own writes immediately should read from the primary.
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.
