Change the data replication mode

ApsaraDB RDS for PostgreSQL supports three data replication modes for High-availability Edition instances: asynchronous, synchronous, and semi-synchronous. Change the replication mode to balance write performance against data protection based on your workload requirements.

How it works

In a High-availability Edition deployment, your RDS instance continuously replicates data from the primary instance to a standby instance. The data replication mode controls how the primary instance handles transaction commits relative to standby replication:

Asynchronous -- The primary instance commits transactions without waiting for the standby to confirm receipt. This delivers the fastest write performance but allows potential data loss during failover.
Synchronous -- The primary instance waits for the standby to confirm that data has been written before committing. This provides the strongest data protection at the cost of higher write latency.
Semi-synchronous -- The primary instance waits for the standby up to a configurable timeout. If the standby does not respond within the timeout, the instance falls back to asynchronous mode temporarily. After data catches up, it returns to semi-synchronous mode automatically. This balances performance and protection for most production workloads.

When to use each mode

Data replication mode	Protection level	Performance	Data loss risk on failover	Best for
Asynchronous (default)	Optimal performance	Highest write speed	Medium	Logging, analytics ingestion, dev/test workloads where brief data loss is acceptable
Synchronous	Optimal protection	Medium write speed	Lowest	Financial transactions, order processing, workloads requiring strong data protection
Semi-synchronous	High availability	Balanced	Low (bounded by timeout)	Most production workloads that need both reliable performance and strong data protection

Prerequisites

Before you begin, make sure that you have:

An RDS instance running High-availability Edition with cloud disks
Minor engine version 20220228 or later (required only for the Semi-synchronous mode). For details, see Update the minor engine version

Check the current data replication mode

Go to the Instances page. In the top navigation bar, select the region of your RDS instance. Then, click the instance ID.
In the left-side navigation pane, click Service Availability.
In the Instance Availability section, find the Data Replication Mode value.

Method 1: Use the console

Go to the Instances page. In the top navigation bar, select the region of your RDS instance. Then, click the instance ID.
In the left-side navigation pane, click Service Availability.
Click Change Data Replication Mode.
Select the desired Data Replication Mode.

Important
To use Semi-synchronous mode, make sure that the minor engine version of your RDS instance is 20220228 or later. For details, see Update the minor engine version.
Click Ok.

Method 2: Modify parameters

Modify the synchronous_commit, rds_sync_replication_timeout, and synchronous_standby_names parameters directly to change the replication mode.

Go to the Instances page. In the top navigation bar, select the region of your RDS instance. Then, click the instance ID.
In the left-side navigation pane, click Parameters.
On the Modifiable Parameters tab, find the synchronous_commit, rds_sync_replication_timeout, and synchronous_standby_names parameters.
In the Running Value column, click the edit icon next to the parameter value. Enter the new value and click Ok.

Set the parameters according to the target replication mode:

Asynchronous mode

Parameter Value

synchronous_commit off

Note
When synchronous_commit is set to off, the rds_sync_replication_timeout parameter has no effect.

Synchronous mode

Parameter Value

synchronous_commit remote_write

rds_sync_replication_timeout 0

synchronous_standby_names 'standby1'

Semi-synchronous mode

Parameter Value

synchronous_commit remote_write

rds_sync_replication_timeout Greater than 0. Range: 0--300000 ms. Recommended: 500.

synchronous_standby_names 'standby1'

Note
The rds_sync_replication_timeout parameter specifies the synchronization timeout in milliseconds. If a synchronization operation times out, the instance temporarily falls back to asynchronous mode (optimal performance protection level). After data synchronization catches up, the instance automatically returns to semi-synchronous mode (high availability protection level).

Important
The default value of synchronous_standby_names is standby1. If you have a self-managed secondary instance, do not name it standby1. Otherwise, the RDS instance may replicate data to the self-managed instance in synchronous or semi-synchronous mode, which can cause data loss during a primary/secondary switchover.
Click Apply Changes.

Parameter	Value
`synchronous_commit`	`off`

Parameter	Value
`synchronous_commit`	`remote_write`
`rds_sync_replication_timeout`	`0`
`synchronous_standby_names`	`'standby1'`

Parameter	Value
`synchronous_commit`	`remote_write`
`rds_sync_replication_timeout`	Greater than `0`. Range: 0--300000 ms. Recommended: `500`.
`synchronous_standby_names`	`'standby1'`

Verify the change

After changing the data replication mode, verify that the new mode is active:

In the left-side navigation pane, click Service Availability.
In the Instance Availability section, confirm that the Data Replication Mode value matches your configuration.

Usage notes

This feature is available only for instances running High-availability Edition with cloud disks.
Semi-synchronous mode requires minor engine version 20220228 or later.
If you have a self-managed secondary instance, do not name it standby1. The default value of synchronous_standby_names is standby1. A naming conflict can cause the RDS instance to replicate data to the self-managed instance in synchronous or semi-synchronous mode, leading to potential data loss during a primary/secondary switchover.
In semi-synchronous mode, if the standby does not respond within the rds_sync_replication_timeout period, the instance temporarily operates in asynchronous mode. Normal semi-synchronous replication resumes automatically after the standby catches up.

References

Update the minor engine version