Manually upgrade or downgrade the CPU and memory specifications of compute nodes in a PolarDB cluster to meet performance requirements or control costs as your workload changes.
Potential impacts
A specification change triggers a node restart, which causes transient disconnections and brief performance fluctuations. The exact impact depends on your cluster's hot standby support and binary logging configuration.
| Cluster condition | Impact |
|---|---|
| No failover with hot standby | Transient disconnection of about 20–30 seconds. Make sure your application has a reconnection mechanism before you start. |
| Failover with hot standby + binary logging enabled | Transient disconnection of about 5–10 seconds. |
| Failover with hot standby + binary logging disabled | Performance drops to zero for 1–3 seconds. Connections and transactions are not interrupted. |
Keep the following in mind before you start:
-
Change duration: The task takes about 5 minutes per compute node. For example, changing two nodes takes about 10 minutes. Actual duration depends on cluster workload and the number of databases and tables.
-
Data latency: Data synchronization latency on read-only nodes may be higher than usual during the change.
-
Data security: The change only adjusts compute node resources and does not affect existing data.
-
Restart order: When upgrading or downgrading the primary and read-only nodes at the same time, the system restarts read-only nodes first, then the primary node. No primary/standby switchover occurs.
Node specification limits
To maintain cluster stability and high availability, the following limits apply:
-
At least one read-only node must have the same specifications as the primary node.
-
A read-only node with hot standby enabled must have the same specifications as the primary node.
-
The memory of a read-only node cannot be less than half the memory of the primary node.
-
Minimum CPU cores for read-only nodes based on the primary node:
| Primary node CPU | Minimum read-only node CPU |
|---|---|
| 2 | 2 |
| 4 | 4 |
| 8 | 4 |
| 16 | 8 |
| 32 | 16 |
| 64 | 32 |
| 88 | 64 |
| 120 | 64 |
These limits do not apply to Multi-master Cluster (Limitless) Edition clusters. Because of their analytical workload characteristics, In-Memory Column Index (IMCI) read-only nodes should have higher specifications than the primary node.
Prerequisites
Before you begin, ensure that you have:
-
A Subscription or Pay-as-you-go cluster
-
A cluster in the Running state with no other configuration change tasks in progress (such as adding nodes, deleting nodes, or performing minor version upgrades)
Change compute node specifications
-
Log on to the PolarDB console. In the left navigation pane, click Clusters and select the region where your cluster is located.
-
Go to the Change Configurations page using one of these methods:
-
Click Change Configurations in the Actions column of the target cluster.
-
Click the cluster ID to go to the Basic Information page. In the Database Nodes section, click Change Configurations.
-
-
In the Change Configuration dialog box, select Upgrade or Downgrade, then click OK.
-
On the Upgrade/Downgrade page, review Current Configuration and configure the following parameters:
Parameter Description Sub-series Select the target specification series: General Specification or Dedicated Specification. Node Select new specifications for the nodes to change. Set specifications for the entire cluster or configure each node individually. At least one read-only node must have the same specifications as the primary node. Switchover time Switch Now: The change takes effect immediately. Scheduled Switch: Select a time within the next 24 hours. The change completes within 30 minutes after the selected time. View or cancel the task on the Scheduled Tasks page. -
Read and agree to the Terms of Service, click Buy Now, and complete the payment. After the payment is successful, the task executes at the scheduled switchover time.
Billing
Changing specifications affects billing based on the cluster's billing method:
-
Pay-as-you-go: After a successful change, you are charged hourly at the new specification price.
-
Subscription upgrade: Pay the price difference between the new and old specifications for the remaining subscription period.
-
Subscription downgrade: The system calculates the remaining value and issues a refund. For more information, see Refund amount calculation.
API reference
| API | Description |
|---|---|
| ModifyDBNodeClass | Changes the specifications of a PolarDB cluster node. |
| ModifyDBNodesClass | Changes the specifications of a single node in a PolarDB cluster. |
FAQ
What do I do if I get an "Insufficient Resources" error when selecting specifications?
The Insufficient Resources. Please Change The Region Or Zone And Try Again. message means the selected specifications are not available in the current zone. Try one of the following:
-
Select similar specifications that are slightly lower or higher.
-
Manually change the zone to avoid similar resource shortages in the future.
How long does the specification change take?
The change takes about 5 minutes per compute node. A two-node cluster takes about 10 minutes. Actual duration varies with cluster workload and the number of databases and tables.
Can I cancel a scheduled specification change?
Yes. Go to the Scheduled Tasks page to view or cancel a pending task.
Will read-only node data lag during the change?
Yes. Data synchronization latency on read-only nodes may be higher than usual while the specification change is in progress. Latency returns to normal after the change completes.