RDS DuckDB-based analytical primary instances support five categories of configuration changes: instance type and storage capacity, node configuration, automatic storage expansion, primary/secondary switchover policy, and billing method.
Before making any changes in production, test them on a non-production instance to understand the operational impact.
Configuration orders cannot be canceled after submission. Evaluate your requirements before proceeding.
At a glance: downtime and scope
| Change | Downtime | Affected scope |
|---|---|---|
| Instance type (specs) | ~30 seconds (switchover) | Entire instance |
| Storage expansion | None (in most cases) | Entire instance |
| Add node | None | New node only |
| Delete node | Brief interruption on deleted node | Deleted node only |
| Change secondary node specs | Brief interruption on secondary node | Secondary node only |
| Change primary node specs | ~30-second switchover | Primary + involved secondary |
| Automatic storage expansion | None | Entire instance |
| Primary/secondary switchover | Up to 30 seconds | Entire instance |
| Billing method change | None | Entire instance |
Change the instance type and storage capacity
Change the instance type to upgrade or downgrade computing specifications. Storage capacity can only be expanded — reduction is not currently supported.
Prerequisites
Before you begin, make sure that:
Your Alibaba Cloud account has no unpaid renewal orders
The instance status is Running (if locked, unlock the instance first)
Potential impacts
| Change | Impact |
|---|---|
| Specifications | Takes several minutes (not affected by data volume). An instance switchover occurs, which takes about 30 seconds. Make the change during off-peak hours and configure an automatic reconnection mechanism in your application. |
| Storage expansion | No transient disconnections in most cases. If a snapshot task is running, expansion starts after the snapshot completes. |
| Storage reduction | Not supported. |
The specification change does not cause data loss or change the instance ID and endpoint. No manual restart is required.
Avoid DDL operations from the time you submit the order until the change is complete.
Billing
The actual fee is subject to the price displayed in the console. For details, see Billing rules for specification changes.
Change instance specifications or storage
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the Configuration Information section of the Basic Information page, click Change Specifications.
(Subscription instances only) In the dialog box, select a change method and click Next.
Upgrade or Downgrade: The new configuration takes effect immediately. After the change task is submitted, the system syncs disk data to a new instance. At the scheduled switchover time, the system transfers the instance ID and endpoint to the new instance — both remain unchanged.
On the Change Specifications page, select the new instance type. The current instance type is selected by default.
Select a Switching Time:
Execute Immediately: Switches immediately after data migration completes.
Switch Within Maintenance Window: Switches within the configured maintenance window.
ImportantIf you select Switch Within Maintenance Window, the instance status stays Upgrading/Downgrading until the switchover completes. During this period, you cannot upgrade or downgrade specs, upgrade the database engine version, or migrate the instance across zones. For storage-only expansions, business is not affected in most cases and the change takes effect immediately — you do not need to select Switch Within Maintenance Window.
On the Change Instance page, confirm the before-and-after details, click Confirm Order, and complete payment.
FAQ
My instance disk was expanded but the instance is still locked. Why?
If an instance is locked because its disk is full, it unlocks automatically after the expansion task completes. Click Task Center in the upper-right corner of the Basic Information page to check progress.
Why does a storage expansion cause a transient disconnection?
A storage expansion requires an instance switchover. For more information about the impact of transient disconnections, see Potential impacts above.
Does the free backup quota increase when I expand disk storage?
Yes. For details, see Free backup quota.
How long is the service unavailable when I upgrade CPU, memory, and disk at the same time?
The service is unavailable for several minutes regardless of whether you change one item or multiple. During the upgrade, an instance switchover or restart may occur, and most database, account, and network operations cannot be performed. Schedule the change during the maintenance window. For impact details, see Potential impacts above.
What factors affect the duration of a specification change?
Specification changes for RDS DuckDB-based analytical primary instances that use disk storage do not involve cross-machine migration. Therefore, the change duration is short. For details, see Potential impacts above.
Does the endpoint change after a configuration change?
No. The endpoint (for example, rm-bpxxxxx.mysql.rds.aliyuncs.com) stays the same, but the underlying IP address may change. Use the endpoint rather than the IP address in your application.
How do I configure an automatic reconnection mechanism?
For Java applications, set the Time to Live (TTL) to no more than 60 seconds. This allows the application to pick up the new virtual IP address (VIP) after a DNS re-query. See the official JDK documentation for how to configure TTL in Java.
Change node configurations
RDS DuckDB-based analytical primary instances support adding and deleting cluster nodes, and independently changing node specifications and storage capacity.
Add a node
During node addition, read and write operations on the instance are not affected. After the node is added, configure the cluster read-only endpoint weights to distribute traffic to the new node for load balancing.
A single instance supports up to eight secondary nodes.
Prerequisites
The instance must be in the Running state.
Billing
The actual fee is subject to the price displayed in the console. For details, see Billing rules for specification changes.
Add a node
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the Instance Topology Management section of the Basic Information page, click Add or Delete Node.
In the dialog box, select Add Node and click Next.
In the Node section, click the
To cancel adding a node, click the
icon to the left of the node.Read the Terms of Service and click Confirm Order.
The node is added when the instance status changes from Adding Node to Running.
What's next
Adjust the node configuration of the cluster read-only endpoint to route traffic to the new node.
Delete a node
Deleting a node causes a brief service interruption on that node only — other nodes are not affected. A Cluster Edition instance must always retain at least one primary node and one secondary node.
Before deleting a node, set its read weight to 0 to avoid routing traffic to it during the deletion process.
Prerequisites
The instance must be in the Running state.
Billing
The actual fee is subject to the price displayed in the console. For details, see Billing rules for specification changes.
Delete a node
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the Instance Topology Management section of the Basic Information page, click Add or Delete Node.
In the dialog box, select Delete Node and click Next.
To the left of the node you want to delete, click the
icon.Read the Terms of Service and click Confirm Order.
What's next
After deletion, adjust the cluster read-only endpoint node configuration to redistribute traffic to the remaining nodes.
Change node specifications and storage capacity
DuckDB-based analytical primary instances support different specifications per cluster node and independent storage type and capacity adjustments per node, giving you fine-grained control over cost and performance.
Prerequisites
Before you begin, make sure that:
The instance status is Running
The data replication relationship between primary and secondary nodes is semi-synchronous or asynchronous
Limits
| Change | Scope | Constraints |
|---|---|---|
| Node specifications | Single node or entire instance | At least one node must have the same specifications as the primary node. All nodes must share the same product type and instance family. |
| Node storage capacity | Entire instance only | Expansion: no transient disconnections (if a snapshot task is running, expansion starts after it completes). Reduction: not supported. |
View node specifications in the topology diagram at the bottom of the Basic Information page.
Potential impacts
| Change | Impact |
|---|---|
| Secondary node specs | The secondary node experiences a brief service interruption. The primary node is not affected. |
| Primary node specs | A primary/secondary switchover is triggered. The primary node and the involved secondary node experience a brief service interruption. |
The specification change takes several minutes and is not affected by data volume. During an instance switchover, most database, account, and network operations cannot be performed. Schedule changes during off-peak hours and configure an automatic reconnection mechanism in your application.
Node IDs and endpoints do not change after a specification change. No manual restart is required.
For details on switchover impact, see Impacts of an instance switchover.
Billing
The actual fee is subject to the price displayed in the console. For details, see Billing rules for specification changes.
Change node specifications or storage
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the Configuration Information or Instance Topology Management section of the Basic Information page, click Specification Change.
Select the Storage Type, Storage capacity, Node Change, and Switching Time as needed.
Read the Terms of Service, click Confirm Order, confirm the change details, and click Continue to complete payment.
The change takes several minutes. The instance status shows Upgrading/Downgrading during the change and returns to Running when complete.
Configure automatic storage expansion
If an instance runs out of storage, write operations fail — which can cause data loss or a database crash. Automatic storage expansion scales out the instance when storage usage reaches a configured threshold, with no restart and no business interruption.
After storage is scaled out, scale-in is not supported automatically. To reduce storage, use the manual process described in Change the instance type and storage capacity.
Prerequisites
Before you begin, make sure that:
The instance is in the Running state
Your account has a sufficient balance to cover the expansion cost
Enable and configure automatic storage expansion
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
On the Basic Information page, in the Instance Resources section, click Settings next to Automatic Storage Expansion.
If Settings is not displayed, the instance does not meet the prerequisites.
Configure the following parameters.
Parameter Description Automatic Storage Expansion Enable or disable automatic storage expansion. Available Storage ≤ The free-space threshold that triggers expansion. When free space drops to or below this percentage, expansion is triggered. The amount added is the greater of: 5 GB (or 10 GB if total storage is less than 50 GB and free space is below 10%), or 15% of current storage capacity rounded to the nearest multiple of 5. For example, on a 100 GB instance, 15% is 15 GB — so 15 GB is added, bringing the total to 115 GB. Maximum Storage Capacity The maximum total storage capacity after expansion. Must be greater than or equal to the current total. Maximum value: 64 TB. Click OK.
Configure primary/secondary switchover
RDS automatically triggers a primary/secondary switchover if the primary node fails or if an emergency fix is applied to the secondary node. During a switchover, the primary and secondary node roles are swapped. The instance endpoint stays the same, and your application automatically connects to the new primary node.
How it works
Automatic switchover (enabled by default): If the primary node fails and becomes inaccessible, the system switches to the secondary node. For what triggers a switchover, see Reasons for a primary/secondary switchover.
Manual switchover: Available even when automatic switchover is enabled. Use manual switchovers for disaster recovery drills or to connect to the nearest node in a multi-zone deployment.
Potential impacts
The service may be unavailable for up to 30 seconds during a switchover. Configure an automatic reconnection mechanism in your application. If your application uses an older version of the Druid connection pool, connections may not recover automatically after a disconnection. Upgrade Druid to version 1.1.16 or later to resolve this.
The instance endpoint does not change, but the underlying IP address may change. Use the endpoint rather than the IP address in your application.
If the instance is in a failed state at the time of the switchover, the switchover may take longer.
Switch primary and secondary nodes manually
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the left navigation pane, click Service Availability.
In the Instance Availability section, click Switch Primary/Secondary Instance.
In the Primary/Secondary Switchover dialog box, on the Secondary Node row, select the target secondary node to promote to primary.
Select a switchover time and click OK.
ImportantMost operations — such as managing databases and accounts or switching network types — cannot be performed during the switchover. Select Within Specified Time Range to minimize disruption.
Temporarily disable automatic switchover
Automatic switchover is enabled by default. Disable it temporarily in scenarios such as:
Sales promotions or high-traffic events, to prevent an unexpected switchover from affecting availability
Major application upgrades, to avoid introducing variables from a switchover
Critical stability assurance periods
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the left navigation pane, click Service Availability.
In the Instance Availability section, click Automatic Primary/Secondary Switchover.
Select Temporarily Disable, set the Deadline, and click OK.
Automatic switchover resumes after the deadline. By default, the feature is disabled for one day. The maximum end time is 23:59:59 seven days from now.
After saving the settings, the end time for the temporary disablement is shown on the Service Availability page.
View switchover logs
Go to the Instances page. In the top navigation bar, select the region where the instance resides, then click the instance ID.
In the left navigation pane, click Service Availability.
In the Primary/Secondary Switchover Logs section, adjust the time range to view the logs.
FAQ
Do I need to manually switch back after a primary/secondary switchover?
No. The data on the primary and secondary nodes is identical. After a switchover, the original secondary becomes the new primary. No further action is required.
The instance does not recover after a switchover. What should I do?
Persistent connections from your application may not detect the change in connection status after an exception-triggered switchover. Without a socket timeout, the application waits indefinitely for a response, causing SQL execution failures. Configure connectTimeout and socketTimeout to prevent this:
For online transaction scenarios: set
connectTimeoutto 1–2 seconds andsocketTimeoutto 60–90 seconds.
These values are guidelines — adjust them based on your business payload and usage patterns.
Change the billing method
Change from pay-as-you-go to subscription
Prerequisites
Before you begin, make sure that:
The instance is in the Running state
The instance has no unpaid orders to change to the subscription billing method
Usage notes
If you upgrade the specifications of an instance that has an unpaid subscription-change order, that order becomes invalid. Cancel it on the Order Management page before retrying the billing method change.
If the billing method change crosses the top of an hour, the previous hour is billed as pay-as-you-go, and subscription billing starts at the top of the next hour. For example, if you start the change at 09:55 and it completes at 10:10, the period from 09:00 to 10:00 is billed as pay-as-you-go, and subscription billing starts after 10:00.
Pay-as-you-go resources are billed hourly and bills may be delayed. To avoid paying for an extra hour, initiate the change 15–30 minutes before the top of the hour.
Switch to subscription billing
Log on to the RDS console. In the left navigation pane, click Instances. In the top navigation bar, select the region.
Find the target instance. Go to the Change from Pay-as-you-go to Subscription page in one of the following ways:
In the Billing Method column, click Switch To Subscription.
Click the instance ID. On the Basic Information page, in the Status section, find Billing Method and click Change To Subscription.
Select a Subscription Duration and click Confirm Order.
Submitting this order blocks new instance purchases and billing method changes on other instances until the order is paid or canceled. Manage the order on the Order Management page.
Change from subscription to pay-as-you-go
Prerequisites
The instance must be in the Running state.
If a subscription instance is locked because it has expired, renew it first. See Manual renewal.
Fees
After switching to pay-as-you-go, the remaining balance is refunded to your original payment method using the following formula:
Refund amount = Actual amount paid - Used resource fee
Actual amount paid: The cash amount paid for the order. This does not include deductions from vouchers or coupons.
Used resource fee: Daily subscription price (original monthly price / 30) × Actual usage duration × Discount for the used duration
NoteThe actual usage duration is measured in days. A partial day is counted as a full day.
Actual amount paid: The cash paid for the order, excluding vouchers and coupons.
Used resource fee: Daily price (original monthly price / 30) x Actual usage days x Discount for the used duration
A partial day is counted as a full day.
For the specific refund amount, see Standard refund instructions.
Switch to pay-as-you-go billing
Log on to the RDS console. In the left navigation pane, click Instances. In the top navigation bar, select a region.
Find the target instance. Go to the Change from Subscription to Pay-as-you-go page in one of the following ways:
In the Billing Method column, click Switch To Pay-As-You-Go.
Click the instance ID. On the Basic Information page, in the Status section, find Billing Method and click Change to Pay-As-You-Go.
Confirm the instance information, read the Terms of Service, and click Confirm Order.