PolarDB:Manage the serverless feature for a cluster with defined specifications

How scaling works

PolarDB monitors CPU utilization, memory utilization, and other kernel metrics on each node every monitoring cycle and automatically adjusts resources within the limits you set.

Scale-up is triggered on a node when any of the following conditions is met:

• CPU utilization exceeds 85%

• Memory utilization exceeds 85%

• The read-only node's capacity falls below half the primary node's capacity. For example, if a read-only node has 4 PCUs and the primary node has 10 PCUs, the read-only node scales up to at least 5 PCUs.

Scale-out (adding read-only nodes) is triggered when a read-only node is already at its maximum capacity and CPU or memory utilization still exceeds 85%.

Scale-down is triggered when CPU utilization drops below 55% and memory utilization drops below 40%.

Prerequisites

Before you begin, ensure that you have:

• Enabled the serverless feature for a cluster with defined specifications

Usage notes

• PolarDB uses PolarDB Capacity Units (PCUs) for per-second billing and resource elasticity. One PCU is approximately 1 core and 2 GB of memory. PCUs are added to or removed from nodes dynamically, in increments of 0.5 PCU, within the limits you configure.

• Disabling the serverless feature requires a cluster restart. Perform this operation during off-peak hours to minimize impact on your workload.

Modify serverless configuration

1. Log on to the PolarDB console. In the navigation pane on the left, click Clusters. Select the region where the cluster is located, then click the cluster ID.

2. On the Basic Information page, go to the Database Nodes section and click Serverless Configuration.

   

3. In the Configure Serverless-related Parameters dialog box, set the following parameters.

   Basic Configuration

   Parameter	Description	Valid values
   Minimum Resources for Single Node	Minimum PCUs added per node on top of its defined specifications	0–16
   Maximum Resources for Single Node	Maximum PCUs added per node on top of its defined specifications	0–16
   Minimum Number of Read-only Nodes	Minimum number of read-only nodes the serverless feature can add during scale-out	0–15
   Maximum Number of Read-only Nodes	Maximum number of read-only nodes the serverless feature can add during scale-out	0–15

   Note

   For example, if Minimum Resources for Single Node is 2 and Maximum Resources for Single Node is 8, the node's baseline capacity is its defined specifications plus 2 PCUs (approximately 2 cores and 4 GB of memory). Under high load, the serverless feature can scale the node up to its defined specifications plus 8 PCUs (approximately 8 cores and 16 GB of memory). The number of read-only nodes scales automatically within the range you set.

   Advanced Mode

   Adjust these parameters if the default scaling behavior doesn't match your workload's sensitivity requirements.

   Parameter	Description	Valid values
   Scan Interval	Controls how quickly the serverless feature responds to workload changes. Sensitive mode uses a shorter observation window for faster response, but may cause more frequent scaling—use it for workloads with transient CPU spikes that require rapid reaction.	Standard, Sensitive
   Maximum CPU Resources for Elastic Upgrade	CPU utilization threshold that triggers a scale-up	40%–100%
   Minimum CPU Resources for Elastic Upgrade	CPU utilization threshold that triggers a scale-down	10%–70%

   Note

   The scale-down threshold must be at least 30 percentage points lower than the scale-up threshold. For example, if the scale-up threshold is 80%, the scale-down threshold must be 50% or lower.

4. Click OK.

Disable the serverless feature

When you no longer need the serverless feature, you can disable it. After disabling, all PCUs and read-only nodes added by the serverless feature are released, and the cluster reverts to its defined specifications.

1. Log on to the PolarDB console. In the navigation pane on the left, click Clusters. Select the region where the cluster is located, then click the cluster ID.

2. On the Basic Information page, go to the Database Nodes section and click Disable Serverless.

   

API reference