Working with read replicas

Amazon RDS uses the MariaDB, Microsoft SQL Server, MySQL, Oracle, and PostgreSQL DB engines' built-in replication functionality to create a special type of DB instance called a read replica from a source DB instance. The source DB instance becomes the primary DB instance. Updates made to the primary DB instance are asynchronously copied to the read replica. You can reduce the load on your primary DB instance by routing read queries from your applications to the read replica. Using read replicas, you can elastically scale out beyond the capacity constraints of a single DB instance for read-heavy database workloads.

Note

The information following applies to creating Amazon RDS read replicas either in the same AWS Region as the source DB instance, or in a separate AWS Region. The information following doesn't apply to setting up replication with an instance that is running on an Amazon EC2 instance or that is on-premises.

When you create a read replica, you first specify an existing DB instance as the source. Then Amazon RDS takes a snapshot of the source instance and creates a read-only instance from the snapshot. Amazon RDS then uses the asynchronous replication method for the DB engine to update the read replica whenever there is a change to the primary DB instance. The read replica operates as a DB instance that allows only read-only connections. Applications connect to a read replica the same way they do to any DB instance. Amazon RDS replicates all databases in the source DB instance.

Note

The Oracle DB engine supports replica databases in mounted mode. A mounted replica doesn't accept user connections and so can't serve a read-only workload. The primary use for mounted replicas is cross-Region disaster recovery. For more information, see Working with Oracle replicas for Amazon RDS.

In some cases, a read replica resides in a different AWS Region from its primary DB instance. In these cases, Amazon RDS sets up a secure communications channel between the primary DB instance and the read replica. Amazon RDS establishes any AWS security configurations needed to enable the secure channel, such as adding security group entries. For more information about cross-Region read replicas, see Creating a read replica in a different AWS Region.

You can configure a read replica for a DB instance that also has a standby replica configured for high availability. Replication with the standby replica is synchronous, and the standby replica can't serve read traffic.

For more information about high availability and standby replicas, see High availability (Multi-AZ) for Amazon RDS.

Read replicas are supported by the MariaDB, Microsoft SQL Server, MySQL, Oracle, and PostgreSQL DB engines. In this section, you can find general information about using read replicas with all of these engines. For information about using read replicas with a specific engine, see the following sections:

• Working with MariaDB read replicas

• Working with read replicas for Microsoft SQL Server in Amazon RDS

• Working with MySQL read replicas

• Working with Oracle replicas for Amazon RDS

• Working with PostgreSQL read replicas in Amazon RDS

Overview of Amazon RDS read replicas

Deploying one or more read replicas for a given source DB instance might make sense in a variety of scenarios, including the following:

• Scaling beyond the compute or I/O capacity of a single DB instance for read-heavy database workloads. You can direct this excess read traffic to one or more read replicas.

• Serving read traffic while the source DB instance is unavailable. In some cases, your source DB instance might not be able to take I/O requests, for example due to I/O suspension for backups or scheduled maintenance. In these cases, you can direct read traffic to your read replicas. For this use case, keep in mind that the data on the read replica might be "stale" because the source DB instance is unavailable.

• Business reporting or data warehousing scenarios where you might want business reporting queries to run against a read replica, rather than your production DB instance.

• Implementing disaster recovery. You can promote a read replica to a standalone instance as a disaster recovery solution if the primary DB instance fails.

By default, a read replica is created with the same storage type as the source DB instance. However, you can create a read replica that has a different storage type from the source DB instance based on the options listed in the following table.

Source DB instance storage type	Source DB instance storage allocation	Read replica storage type options
PIOPS	100 GiB–32 TiB	PIOPS, GP2, Standard
GP2	100 GiB–32 TiB	PIOPS, GP2, Standard
GP2	<100 GiB	GP2, Standard
Standard	100 GiB–6 TiB	PIOPS, GP2, Standard
Standard	<100 GiB	GP2, Standard

Note

When you increase the allocated storage of a read replica, it must be by at least 10 percent. If you try to increase the value by less than 10 percent, you get an error.

Amazon RDS doesn't support circular replication. You can't configure a DB instance to serve as a replication source for an existing DB instance. You can only create a new read replica from an existing DB instance. For example, if MyDBInstance replicates to ReadReplica1, you can't configure ReadReplica1 to replicate back to MyDBInstance. For MariaDB and MySQL you can create a read replica from an existing read replica. For example, from ReadReplica1, you can create a new read replica, such as ReadReplica2. For Oracle, PostgreSQL, and SQL Server, you can't create a read replica from an existing read replica.

If you no longer need read replicas, you can explicitly delete them using the same mechanisms for deleting a DB instance. If you delete a source DB instance without deleting its read replicas in the same AWS Region, each read replica is promoted to a standalone DB instance. For information about deleting a DB instance, see Deleting a DB instance. For information about read replica promotion, see Promoting a read replica to be a standalone DB instance. If you have cross-Region read replicas, see Cross-Region replication considerations for considerations related to deleting the source for a cross-Region read replica.

Differences between read replicas for different DB engines

Because Amazon RDS DB engines implement replication differently, there are several significant differences you should know about, as shown in the following table.

Feature or behavior	MySQL and MariaDB	Oracle	PostgreSQL	SQL Server
What is the replication method?	Logical replication.	Physical replication.	Physical replication.	Physical replication.
How are transaction logs purged?	RDS for MySQL and RDS for MariaDB keep any binary logs that haven't been applied.	If a primary DB instance has no cross-Region read replicas, Amazon RDS for Oracle keeps a minimum of two hours of transaction logs on the source DB instance. Logs are purged from the source DB instance after two hours or after the archive log retention hours setting has passed, whichever is longer. Logs are purged from the read replica after the archive log retention hours setting has passed only if they have been successfully applied to the database. In some cases, a primary DB instance might have one or more cross-Region read replicas. If so, Amazon RDS for Oracle keeps the transaction logs on the source DB instance until they have been transmitted and applied to all cross-Region read replicas. For information about setting archive log retention hours, see Retaining archived redo logs.	PostgreSQL has the parameter wal_keep_segments that dictates how many write ahead log (WAL) files are kept to provide data to the read replicas. The parameter value specifies the number of logs to keep.	The Virtual Log File (VLF) of the transaction log file on the primary replica can be truncated after it is no longer required for the secondary replicas. The VLF can only be marked as inactive when the log records have been hardened in the replicas. Regardless of how fast the disk subsystems are in the primary replica, the transaction log will keep the VLFs until the slowest replica has hardened it.
Can a replica be made writable?	Yes. You can enable the MySQL or MariaDB read replica to be writable.	No. An Oracle read replica is a physical copy, and Oracle doesn't allow for writes in a read replica. You can promote the read replica to make it writable. The promoted read replica has the replicated data to the point when the request was made to promote it.	No. A PostgreSQL read replica is a physical copy, and PostgreSQL doesn't allow for a read replica to be made writable.	No. A SQL Server read replica is a physical copy and also doesn't allow for writes. You can promote the read replica to make it writable. The promoted read replica has the replicated data up to the point when the request was made to promote it.
Can backups be performed on the replica?	Yes. You can enable automatic backups on a MySQL or MariaDB read replica.	No. You can't create manual snapshots of Amazon RDS for Oracle read replicas or enable automatic backups for them.	Yes, you can create a manual snapshot of a PostgreSQL read replica, but you can't enable automatic backups.	No. You can't create manual snapshots of Amazon RDS for SQL Server read replicas or enable automatic backups for them.
Can you use parallel replication?	Yes. MySQL version 5.6 and later and all supported MariaDB versions allow for parallel replication threads.	Yes. Redo log data is always transmitted in parallel from the primary database to all of its read replicas.	No. PostgreSQL has a single process handling replication.	Yes. Redo log data is always transmitted in parallel from the primary database to all of its read replicas.
Can you maintain a replica in a mounted rather than a read-only state?	No.	Yes. The primary use for mounted replicas is cross-Region disaster recovery. An Active Data Guard license isn't required for mounted replicas. For more information, see Working with Oracle replicas for Amazon RDS.	No.	No.

Creating a read replica

You can create a read replica from an existing DB instance using the AWS Management Console, AWS CLI, or RDS API. You create a read replica by specifying SourceDBInstanceIdentifier, which is the DB instance identifier of the source DB instance that you want to replicate from.

When you create a read replica, Amazon RDS takes a DB snapshot of your source DB instance and begins replication. As a result, you experience a brief I/O suspension on your source DB instance while the DB snapshot occurs.

Note

The I/O suspension typically lasts about one minute. You can avoid the I/O suspension if the source DB instance is a Multi-AZ deployment, because in that case the snapshot is taken from the secondary DB instance.

An active, long-running transaction can slow the process of creating the read replica. We recommend that you wait for long-running transactions to complete before creating a read replica. If you create multiple read replicas in parallel from the same source DB instance, Amazon RDS takes only one snapshot at the start of the first create action.

When creating a read replica, there are a few things to consider. First, you must enable automatic backups on the source DB instance by setting the backup retention period to a value other than 0. This requirement also applies to a read replica that is the source DB instance for another read replica. For MySQL DB instances, automatic backups are supported only for read replicas running MySQL 5.6 and later, but not for MySQL versions 5.5. To enable automatic backups on an RDS for MySQL version 5.6 and later read replica, first create the read replica, then modify the read replica to enable automatic backups.

Note

Within an AWS Region, we strongly recommend that you create all read replicas in the same virtual private cloud (VPC) based on Amazon VPC as the source DB instance. If you create a read replica in a different VPC from the source DB instance, classless inter-domain routing (CIDR) ranges can overlap between the replica and the RDS system. CIDR overlap makes the replica unstable, which can negatively impact applications connecting to it. If you receive an error when creating the read replica, choose a different destination DB subnet group. For more information, see Working with a DB instance in a VPC.

You can't create a read replica in a different AWS account from the source DB instance.

Console

To create a read replica from a source DB instance

1. Sign in to the AWS Management Console and open the Amazon RDS console at https://console.aws.amazon.com/rds/.

2. In the navigation pane, choose Databases.

3. Choose the DB instance that you want to use as the source for a read replica.

4. For Actions, choose Create read replica.

5. For DB instance identifier, enter a name for the read replica.

6. Choose your instance specifications. We recommend that you use the same DB instance class and storage type as the source DB instance for the read replica.

7. For Multi-AZ deployment, choose Yes to create a standby of your replica in another Availability Zone for failover support for the replica.

   Note

   Creating your read replica as a Multi-AZ DB instance is independent of whether the source database is a Multi-AZ DB instance.

8. To create an encrypted read replica:

   1. Choose Enable encryption.

   2. For Master key, choose the AWS Key Management Service (AWS KMS) key identifier of the customer master key (CMK).

   Note

   The source DB instance must be encrypted. To learn more about encrypting the source DB instance, see Encrypting Amazon RDS resources.

9. Choose other options, such as storage autoscaling.

10. Choose Create read replica.

AWS CLI

To create a read replica from a source DB instance, use the AWS CLI command create-db-instance-read-replica. This example also enables storage autoscaling.

Example

For Linux, macOS, or Unix:

aws rds create-db-instance-read-replica \
    --db-instance-identifier myreadreplica \
    --source-db-instance-identifier mydbinstance \
    --max-allocated-storage 1000

For Windows:

aws rds create-db-instance-read-replica ^
    --db-instance-identifier myreadreplica ^
    --source-db-instance-identifier mydbinstance ^
    --max-allocated-storage 1000

RDS API

To create a read replica from a source MySQL, MariaDB, Oracle, PostgreSQL, or SQL Server DB instance, call the Amazon RDS API CreateDBInstanceReadReplica operation with the following required parameters:

• DBInstanceIdentifier

• SourceDBInstanceIdentifier

Promoting a read replica to be a standalone DB instance

You can promote a read replica into a standalone DB instance. When you promote a read replica, the DB instance is rebooted before it becomes available.

There are several reasons you might want to promote a read replica to a standalone DB instance:

• Performing DDL operations (MySQL and MariaDB only) – DDL operations, such as creating or rebuilding indexes, can take time and impose a significant performance penalty on your DB instance. You can perform these operations on a MySQL or MariaDB read replica once the read replica is in sync with its primary DB instance. Then you can promote the read replica and direct your applications to use the promoted instance.

• Sharding – Sharding embodies the "share-nothing" architecture and essentially involves breaking a large database into several smaller databases. One common way to split a database is splitting tables that are not joined in the same query onto different hosts. Another method is duplicating a table across multiple hosts and then using a hashing algorithm to determine which host receives a given update. You can create read replicas corresponding to each of your shards (smaller databases) and promote them when you decide to convert them into standalone shards. You can then carve out the key space (if you are splitting rows) or distribution of tables for each of the shards depending on your requirements.

• Implementing failure recovery – You can use read replica promotion as a data recovery scheme if the primary DB instance fails. This approach complements synchronous replication, automatic failure detection, and failover.

  If you are aware of the ramifications and limitations of asynchronous replication and you still want to use read replica promotion for data recovery, you can. To do this, first create a read replica and then monitor the primary DB instance for failures. In the event of a failure, do the following:

  1. Promote the read replica.

  2. Direct database traffic to the promoted DB instance.

  3. Create a replacement read replica with the promoted DB instance as its source.

When you promote a read replica, the new DB instance that is created retains the option group and the parameter group of the former read replica. The promotion process can take several minutes or longer to complete, depending on the size of the read replica. After you promote the read replica to a new DB instance, it's just like any other DB instance. For example, you can create read replicas from the new DB instance and perform point-in-time restore operations. Because the promoted DB instance is no longer a read replica, you can't use it as a replication target. If a source DB instance has several read replicas, promoting one of the read replicas to a DB instance has no effect on the other replicas.

Backup duration is a function of the number of changes to the database since the previous backup. If you plan to promote a read replica to a standalone instance, we recommend that you enable backups and complete at least one backup prior to promotion. In addition, you can't promote a read replica to a standalone instance when it has the backing-up status. If you have enabled backups on your read replica, configure the automated backup window so that daily backups don't interfere with read replica promotion.

The following steps show the general process for promoting a read replica to a DB instance:

1. Stop any transactions from being written to the primary DB instance, and then wait for all updates to be made to the read replica. Database updates occur on the read replica after they have occurred on the primary DB instance, and this replication lag can vary significantly. Use the Replica Lag metric to determine when all updates have been made to the read replica.

2. For MySQL and MariaDB only: If you need to make changes to the MySQL or MariaDB read replica, you must set the read_only parameter to 0 in the DB parameter group for the read replica. You can then perform all needed DDL operations, such as creating indexes, on the read replica. Actions taken on the read replica don't affect the performance of the primary DB instance.

3. Promote the read replica by using the Promote option on the Amazon RDS console, the AWS CLI command promote-read-replica, or the PromoteReadReplica Amazon RDS API operation.

   Note

   The promotion process takes a few minutes to complete. When you promote a read replica, replication is stopped and the read replica is rebooted. When the reboot is complete, the read replica is available as a new DB instance.

4. (Optional) Modify the new DB instance to be a Multi-AZ deployment. For more information, see Modifying an Amazon RDS DB instance and High availability (Multi-AZ) for Amazon RDS.

Console

To promote a read replica to a standalone DB instance

1. Sign in to the AWS Management Console and open the Amazon RDS console at https://console.aws.amazon.com/rds/.

2. In the Amazon RDS console, choose Databases.

   The Databases pane appears. Each read replica shows Replica in the Role column.

3. Choose the read replica that you want to promote.

4. For Actions, choose Promote.

5. On the Promote Read Replica page, enter the backup retention period and the backup window for the newly promoted DB instance.

6. When the settings are as you want them, choose Continue.

7. On the acknowledgment page, choose Promote Read Replica.

AWS CLI

To promote a read replica to a standalone DB instance, use the AWS CLI promote-read-replica command.

Example

For Linux, macOS, or Unix:

aws rds promote-read-replica \
    --db-instance-identifier myreadreplica

For Windows:

aws rds promote-read-replica ^
    --db-instance-identifier myreadreplica

RDS API

To promote a read replica to a standalone DB instance, call the Amazon RDS API PromoteReadReplica operation with the required parameter DBInstanceIdentifier.

Monitoring read replication

You can monitor the status of a read replica in several ways. The Amazon RDS console shows the status of a read replica in the Availability and durability section of the read replica details. To view the details for a read replica, choose the name of the read replica in the list of instances in the Amazon RDS console.

You can also see the status of a read replica using the AWS CLI describe-db-instances command or the Amazon RDS API DescribeDBInstances operation.

The status of a read replica can be one of the following:

• replicating – The read replica is replicating successfully.

• replication degraded (SQL Server only) – Replicas are receiving data from the primary instance, but one or more databases might be not getting updates. This can occur, for example, when a replica is in the process of setting up newly created databases.

  The status doesn't transition from replication degraded to error, unless an error occurs during the degraded state.

• error – An error has occurred with the replication. Check the Replication Error field in the Amazon RDS console or the event log to determine the exact error. For more information about troubleshooting a replication error, see Troubleshooting a MySQL read replica problem.

• terminated (MariaDB, MySQL, or PostgreSQL only) – Replication is terminated. This occurs if replication is stopped for more than 30 consecutive days, either manually or due to a replication error. In this case, Amazon RDS terminates replication between the primary DB instance and all read replicas. Amazon RDS does this to prevent increased storage requirements on the source DB instance and long failover times.

  Broken replication can affect storage because the logs can grow in size and number due to the high volume of errors messages being written to the log. Broken replication can also affect failure recovery due to the time Amazon RDS requires to maintain and process the large number of logs during recovery.

• stopped (MariaDB or MySQL only) – Replication has stopped because of a customer-initiated request.

• replication stop point set (MySQL only) – A customer-initiated stop point was set using the mysql.rds_start_replication_until stored procedure and the replication is in progress.

• replication stop point reached (MySQL only) – A customer-initiated stop point was set using the mysql.rds_start_replication_until stored procedure and replication is stopped because the stop point was reached.

Monitoring replication lag

You can monitor replication lag in Amazon CloudWatch by viewing the Amazon RDS ReplicaLag metric.

For MySQL and MariaDB, the ReplicaLag metric reports the value of the Seconds_Behind_Master field of the SHOW SLAVE STATUS command. Common causes for replication lag for MySQL and MariaDB are the following:

• A network outage.

• Writing to tables with indexes on a read replica. If the read_only parameter is not set to 0 on the read replica, it can break replication.

• Using a nontransactional storage engine such as MyISAM. Replication is only supported for the InnoDB storage engine on MySQL and the XtraDB storage engine on MariaDB.

When the ReplicaLag metric reaches 0, the replica has caught up to the primary DB instance. If the ReplicaLag metric returns -1, then replication is currently not active. ReplicaLag = -1 is equivalent to Seconds_Behind_Master = NULL.

For Oracle, the ReplicaLag metric is the sum of the Apply Lag value and the difference between the current time and the apply lag's DATUM_TIME value. The DATUM_TIME value is the last time the read replica received data from its source DB instance. For more information, see V$DATAGUARD_STATS in the Oracle documentation.

For SQL Server, the ReplicaLag metric is the maximum lag of databases that have fallen behind, in seconds. For example, if you have two databases that lag 5 seconds and 10 seconds, respectively, then ReplicaLag is 10 seconds. The ReplicaLag metric returns the value of the following query.

select ag.name name, MAX(hdrs.secondary_lag_seconds) max_lag from sys.dm_hadr_database_replica_state

For more information, see secondary_lag_seconds in the Microsoft documentation.

ReplicaLag returns -1 if RDS can't determine the lag, such as during replica setup, or when the read replica is in the error state.

Note

New databases aren't included in the lag calculation until they are accessible on the read replica.

For PostgreSQL, the ReplicaLag metric returns the value of the following query.

SELECT extract(epoch from now() - pg_last_xact_replay_timestamp()) AS slave_lag

PostgreSQL versions 9.5.2 and later use physical replication slots to manage write ahead log (WAL) retention on the source instance. For each cross-Region read replica instance, Amazon RDS creates a physical replication slot and associates it with the instance. Two Amazon CloudWatch metrics, Oldest Replication Slot Lag and Transaction Logs Disk Usage, show how far behind the most lagging replica is in terms of WAL data received and how much storage is being used for WAL data. The Transaction Logs Disk Usage value can substantially increase when a cross-Region read replica is lagging significantly.

For more information about monitoring a DB instance with CloudWatch, see Monitoring Amazon RDS metrics with Amazon CloudWatch.