Change Tair instance configuration - architecture, memory, shards, replicas - Tair (Redis® OSS-Compatible)

Tair (Redis OSS-compatible) lets you change instance configurations, including architecture, memory capacity, shard count, and replica count.

Supported configuration changes and impacts

What remains unchanged during a configuration change?

All configuration changes:

Connection endpoints, accounts, passwords, and allowlist remain unchanged. No application code changes required.
Data is typically preserved. In rare cases, such as a primary node failure during switchover, a small amount of unsynchronized data might be lost.

Change type	Impact
Change from standalone to high-availability (master-replica) architecture	One to two transient disconnections, each lasting less than 30 seconds. The instance becomes read-only for about one minute.
Change from Standard (master-replica) to cluster architecture	One to two transient disconnections, each lasting less than 30 seconds. The instance becomes read-only for about one minute.
Change from cluster to Standard (master-replica) architecture	One to two transient disconnections, each lasting less than 30 seconds. The instance becomes read-only for about one minute.
Scale the instance specification up or down	May cause one to two transient disconnections, each lasting less than 30 seconds. The instance becomes read-only for about one minute.
Adjust the number of cluster shards	May cause one to two transient disconnections, each lasting less than 30 seconds. The instance becomes read-only for about one minute.
Enable or disable read/write splitting	A transient disconnection that lasts for a few seconds.
Add or remove replica nodes	None.
Change to the cloud-native deployment mode	One transient disconnection lasting less than 30 seconds. The instance becomes read-only for about one minute.

FAQ

Check scaling history

In Task Center, filter by Status = Successful and look for tasks named Modify Instance Specification or Instance. Alternatively, call the DescribeHistoryTasks API operation.

Configuration change failures

Large keys can cause configuration changes to fail.
Identify and delete large keys before changing the configuration. Offline full key analysis.
Downgrades require that used memory is less than 80% of the new capacity (Used memory < New capacity × 0.8). For example, an 8 GB Standard (master-replica) memory-based instance using 2 GB can be downgraded to 4 GB.

Change storage medium for Tair (Enterprise Edition)

Tair (Enterprise Edition) does not support changing between different storage media, such as memory-based, persistent memory, and ESSD-based.

Upgrade CPU performance

Tair (and open source Redis) does not support standalone CPU upgrades. To improve CPU performance:

Change a Standard (master-replica) instance to cluster or read/write splitting architecture.
Add read-only nodes for read/write splitting instances.
Add shards for cluster instances.

How do I upgrade the CPU specification of an instance?

Instance specifications: Instance specifications overview and FAQ.

High-availability to standalone

High-availability instances cannot be changed to standalone because standalone instances lack equivalent data reliability.

To use a standalone instance, purchase a new one and migrate data with DTS. Migrate data between Tair (Redis OSS-compatible) instances.

Pausing applications during change

No, but the instance may become read-only for about one minute with one or two brief disconnections (under 30 seconds each). Schedule changes during off-peak hours. Supported configuration changes and impacts.

Automatic data migration

Yes. The system automatically migrates and rebalances data across all shards.

Configuration change duration

Duration depends on network conditions, request volume, and data size, and cannot be estimated in advance.

Monitor progress by clicking the icon in the upper-right corner of the instance details page.

Losing backup sets

No. However, reducing shards for a Classic cluster instance or changing to Standard (master-replica) architecture alters the mapping between historical backup sets and nodes.

Find historical backup sets by creation time or backup set ID.

To restore data, download the backup set (RDB file), parse it, and import the data into a new instance.

Delayed configuration update

This may be due to a metadata cache refresh delay. Wait a few minutes and refresh the page.

Immediate task execution

For cloud-native high-availability Standard instances, scaling is zero-impact when local resources are sufficient — the task runs immediately, bypassing the maintenance window. If local resources are insufficient, the task runs during the maintenance window with one or two brief disconnections (under 30 seconds each).

API reference

API operation	Description
ModifyInstanceSpec	Changes the specification of an instance.