Tair (Redis OSS-compatible) lets you switch an instance between the standard (master-replica) and cluster architectures.
Limitations
Not all instance types support architecture changes. Review the table below before proceeding.
Instance type | Standard to cluster | Cluster to standard |
Standard instance with read/write splitting enabled | Disable read/write splitting first | Disable read/write splitting first |
Child instance of a distributed instance | Not supported | Not supported |
Tair (Enterprise Edition) SSD-based instance | Not supported | Not supported |
Cluster instance in direct connection mode | — | Not supported |
Billing
Charges depend on your billing method:
Pay-as-you-go: Billed at the new specification rate immediately after the change.
Subscription: Charged or refunded the price difference depending on whether you upgrade or downgrade.
For details, see Configuration changes.
Change from standard (master-replica) to cluster
Before you begin
Review these impacts before starting the change.
Endpoints, accounts, passwords, and whitelists remain unchanged. No application code changes are required.
Data is typically preserved. In the rare event that the primary node fails during switchover, a small amount of unsynchronized data may be lost.
1–2 transient disconnections, each under 30 seconds. Make sure your application has a reconnection mechanism.
Read-only state for about 1 minute. The instance enters read-only mode while the new instance syncs incremental data and DNS cache clears. High write-load instances may experience a longer read-only period.
Lua scripts may be lost. Back up your Lua scripts before proceeding. For details, see Special limits on cluster instances.
In proxy mode clusters, the first argument of redis.call or redis.pcall in Lua scripts must be a literal string (for example, 'GET'), not a variable. A variable argument triggers the error ERR bad lua script for redis cluster, first parameter of redis.call/redis.pcall must be a single literal string. This restriction affects third-party frameworks such as Redisson that construct command names dynamically. To disable this check, set the script_check_enable parameter to 0. For more information, see Special limits on cluster instances.
Additional command limitations apply. Some commands are not supported in cluster architecture. Evaluate the impact on your workload before changing. For details, see Command limitations for cluster instances.
Instance is upgraded to the latest minor version. Minor versions are forward-compatible, so no compatibility issues are expected.
Change the architecture
Log on to the console and go to the Instances page. In the top navigation bar, select the region where your instance resides, then click the instance ID.
In the upper-right corner, click Specification Adjustment, then:
For a subscription instance: select Specification Upgrade.
For a pay-as-you-go instance: select Specifications Upgrade/Downgrade.
On the specification change page, select the target configuration, then click Buy Now. For the Switching Time parameter, choose one of the following options:
Option
Behavior
Switch Within Maintenance Window (recommended)
The switchover runs during the maintenance window (off-peak hours). Before the switchover, go to Task Hub and click Modify Switchover Time to adjust the time if needed.
Switch after Data Migration
The switchover runs immediately after data migration completes.
Complete the payment as prompted.
After you submit the request, the instance status changes to Adjusting configuration regardless of the switching time you selected. This status does not affect your running services — the system is preparing resources and syncing data in the background. Transient disconnections only occur at the moment of switchover.
After the change
Connection mode defaults to proxy mode. Monitor client connections on the proxy node monitoring page. The connection count on data nodes is displayed as 0.
Alert settings are disabled. Existing application groups in Cloud Monitor may also be disabled. Reconfigure them to resume monitoring.
Data flashback is disabled. Reconfigure the feature to resume point-in-time recovery.
If your application relies on keyspace notifications (notify-keyspace-events), reconfigure the parameter on the Parameter Settings page after the change completes.
Change from cluster to standard (master-replica)
Before you begin
Review these impacts before starting the change.
Endpoints, accounts, passwords, and whitelists remain unchanged. No application code changes are required.
Data is typically preserved. In the rare event that the primary node fails during switchover, a small amount of unsynchronized data may be lost.
1–2 transient disconnections, each under 30 seconds. Make sure your application has a reconnection mechanism.
Read-only state for about 1 minute. The instance enters read-only mode while the new instance syncs incremental data and DNS cache clears. High write-load instances may experience a longer read-only period.
Instance is upgraded to the latest minor version. Minor versions are forward-compatible, so no compatibility issues are expected.
Change the architecture
Log on to the console and go to the Instances page. In the top navigation bar, select the region where your instance resides, then click the instance ID.
For a subscription instance, in the upper-right corner, click Specification Adjustment and select Specification Downgrade.
ImportantPay-as-you-go instances cannot downgrade from cluster architecture to standard (master-replica) architecture directly from the console. Use one of the following methods instead:
Call the ModifyInstanceSpec API to downgrade the instance. Make sure to specify the correct
InstanceClassfor the target specification family.First Convert to Subscription, and then follow the steps above.
On the specification change page, select the target configuration, then click Buy Now. For the Switching Time parameter, choose one of the following options:
Option
Behavior
Switch During Maintenance Window (recommended)
The switchover runs during the maintenance window (off-peak hours). Before the switchover, go to Task Hub and click Modify Switchover Time to adjust the time if needed.
Switch after Data Migration
The switchover runs immediately after data migration completes.
Complete the payment as prompted.
After you submit the request, the instance status changes to Adjusting configuration regardless of the switching time you selected. This status does not affect your running services — the system is preparing resources and syncing data in the background. Transient disconnections only occur at the moment of switchover.
After the change
Alert settings are disabled. Existing application groups in Cloud Monitor may also be disabled. Reconfigure them to resume monitoring.
Data flashback is disabled. Reconfigure the feature to resume point-in-time recovery.
FAQ
How long does an architecture change take?
The duration depends on network conditions, request volume, and data size, so it cannot be predicted in advance.
Track progress by clicking the 
icon in the upper-right corner of the instance details page.
