# RDS Multi-AZ vs read replica

AWS comparisons · Updated 17 November 2026

**Short answer:** Multi-AZ is for availability: a synchronous standby in another AZ takes over automatically if the primary fails, but a Multi-AZ DB instance standby cannot serve reads. Read replicas are for read scaling and cross-Region copies: they replicate asynchronously and are readable, but do not fail over automatically. Production databases usually need Multi-AZ; add replicas only for read load or disaster recovery.

## Multi-AZ vs read replica compared

| Factor | Multi-AZ DB instance | Read replica |
|---|---|---|
| Purpose | High availability | Read scaling, reporting, DR |
| Replication | Synchronous | Asynchronous (replica lag possible) |
| Readable | No, the standby serves no traffic | Yes, read-only |
| Failover | Automatic, same endpoint | Manual: promote to standalone instance |
| Location | Another AZ, same Region | Same AZ, another AZ or another Region |
| Backups | Taken from the standby to reduce impact on the primary | Optional on the replica |
| Scaling | One standby | Several replicas, created manually (no replica autoscaling in RDS) |

A third option, the Multi-AZ DB cluster (MySQL and PostgreSQL), has a writer and two readable standbys in three AZs with semi-synchronous replication: availability and some read capacity in one deployment.

## When Multi-AZ is the answer

- Any production database where an AZ or instance failure must not mean a manual recovery.
- Planned maintenance and instance-class changes with less downtime, because RDS can apply changes to the standby and fail over.
- You need zero data loss on failover within the Region, which synchronous replication provides.

## When read replicas are the answer

- Read-heavy applications or reporting queries that should not load the primary.
- Cross-Region disaster recovery: promote the replica if the Region fails, accepting some data loss equal to the lag.
- Upgrades and migrations: test against a replica, or promote it as part of a cut-over.

Watch `ReplicaLag`. Applications that read their own writes from a replica will see stale data during lag; send those reads to the primary.

## Cost and failover behaviour

Multi-AZ roughly doubles the instance and storage cost of the database, because the standby is a full instance you cannot use for reads. Each read replica is also a full instance, billed separately, plus data transfer if it is in another Region.

During a Multi-AZ failover the endpoint DNS name moves to the standby, so clients must reconnect; applications that cache DNS for a long time or hold pooled connections without validation see errors for longer than the failover itself. Test it with a reboot with failover in a quiet window and watch how your connection pool behaves.

## Using both together

The common production layout is a Multi-AZ primary plus one or more read replicas: the standby protects availability, the replicas take read load. Replicas can themselves be Multi-AZ if you depend on them. On Aurora the picture is simpler: Aurora replicas are readable and are the failover targets, so one mechanism serves both purposes.

## Common wrong reasons

- **"A read replica is my high availability."** It will not take over on its own, and promotion breaks replication.
- **"Multi-AZ doubles my read capacity."** The standby in a Multi-AZ DB instance deployment serves no reads.
- **"Multi-AZ protects against bad queries or deletes."** Synchronous replication copies mistakes instantly; use backups and point-in-time recovery for that.

## Questions

### Can I read from an RDS Multi-AZ standby?

Not with a Multi-AZ DB instance deployment. Use a read replica, a Multi-AZ DB cluster (two readable standbys) or Aurora replicas.

### Is a read replica synchronous?

No. RDS read replicas use the engine's asynchronous replication, so they can lag behind the primary.

### Does a read replica fail over automatically?

No. You promote it manually (or with automation), which turns it into a standalone instance.

## When this happens away from your desk

A Multi-AZ failover shows up as a short burst of connection errors and an RDS event; replica lag shows up as stale reads before any alarm. CloudConsole is being built to show RDS events and alarms on a phone with production guardrails.

[About the CloudConsole app](https://cloudconsole.dev/aws-console-app) · [Download on the App Store](https://apps.apple.com/app/cloudconsole-for-aws/id6835419710)

## Sources

- [Multi-AZ DB instance deployments for Amazon RDS](https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Concepts.MultiAZSingleStandby.html)
- [Multi-AZ DB cluster deployments for Amazon RDS](https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/multi-az-db-clusters-concepts.html)
- [Working with DB instance read replicas](https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/USER_ReadRepl.html)

## Related

- [Aurora vs RDS](https://cloudconsole.dev/vs/aurora-vs-rds)
- [Aurora Serverless v2 vs provisioned](https://cloudconsole.dev/vs/aurora-serverless-v2-vs-provisioned)
- [How to fail over an Aurora cluster](https://cloudconsole.dev/how-to/failover-rds-aurora-cluster)
- [CloudWatch alarm: RDS ReplicaLag](https://cloudconsole.dev/alarms/rds-replicalag)
- [Could not connect to RDS: connection timed out](https://cloudconsole.dev/errors/rds-connection-timed-out)
- [How to restore RDS from a snapshot](https://cloudconsole.dev/how-to/restore-rds-from-snapshot)

---

Source: https://cloudconsole.dev/vs/rds-multi-az-vs-read-replica

CloudConsole is an independent iPhone and Android app for Amazon Web Services by Cloudconcur Ltd. It is not affiliated with Amazon Web Services.
