PolarDB:Set a scaling policy for serverless cluster resources

Trigger conditions for scale-up or scale-out

• Scale-up

  PolarDB monitors the CPU utilization, memory usage, and other kernel metrics of the primary node and read-only nodes. During a monitoring cycle, the scale-up of serverless resources is triggered when one of the following conditions is met:

  • When the CPU utilization of a single node is higher than the default scale-up threshold 80% or a custom threshold, the scale-up of the CPU specifications of the node is triggered.

  • When the memory usage of a single node is higher than 90%, the scale-up of the memory specifications of the node is triggered.

  • When the specifications of a read-only node are less than half of the specifications of the primary node, the scale-up of the specifications of the read-only node is triggered. For example, if the specifications of a read-only node are 4 PCUs and the specifications of the primary node are 10 PCUs, the specifications of the read-only node are scaled up to at least 5 PCUs.

• Scale-out

  When a read-only node of a cluster is scaled in to the maximum specifications and the business workloads are still higher than the threshold for a scale-in (CPU utilization is higher than the default threshold 80% or the specified threshold), the scale-out of read-only nodes is triggered.

Trigger conditions for scale-down

When the CPU utilization of a single node is lower than the default scale-down threshold 50% or a custom threshold and the memory usage is lower than 80%, the scale-down of the node is triggered.

Configure current parameters

In the Configure Serverless-related Parameters dialog box, click Edit to the right of the Current Parameters parameter to configure the following parameters:

• Current Parameters

  • Minimum Read-only Nodes: the minimum number of read-only nodes that can be added. Valid values: 0 to 15.

  • Maximum Read-only Nodes: the maximum number of read-only nodes that can be added. Valid values: 0 to 15.

  Note

  • The number of read-only nodes automatically increases or decreases within the specified range based on the actual workloads. For more information, see Scaling.

  • To ensure high availability of serverless clusters, we recommend that you set Minimum Read-only Nodes to 1.

  • Minimum Resources for Single Node: the minimum number of PCUs per node in the cluster. Valid values: 1 PCU to 32 PCUs.

  • Maximum Resources for Single Node: the maximum number of PCUs per node in the cluster. Valid values: 1 PCU to 32 PCUs.

  Note

  Example: If you set the Minimum Resources for Single Node parameter to 2 PCUs and the Maximum Resources for Single Node parameter to 16 PCUs, the default specifications of a node in the serverless cluster are 2 PCUs (2 CPU cores and 4 GB memory). When the business load increases, the system automatically increases the number of PCUs of the primary node or read-only nodes. However, the maximum number of PCUs can only be increased to 16 PCUs (16 CPU cores and 32 GB memory) based on the settings.

  • Read-only Column Store Nodes: the maximum number of read-only column store nodes that can be added. Valid values: 0 to 15.

    Note

    • Before you can add read-only column store nodes to a serverless cluster, make sure that the cluster contains at least one read-only node, which means Minimum Read-only Nodes is set to 1.

    • For more information about read-only column store nodes, see IMCIs.

  • Enable No-activity Suspension: If your serverless cluster does not receive requests within specific periods of time, you can configure the automatic start and stop feature for the instance. After you enable this feature and if no service is connected to the cluster within the period defined by the Detection Period for No-activity Suspension parameter, the cluster automatically enters the suspended state. During the suspension period, you are still charged for the storage space on a pay-as-you-go basis. If any service is connected to the cluster, the cluster immediately starts.

  • Detection Period for No-activity Suspension: The detection period must be a multiple of 5 minutes within a range from 5 minutes to 24 hours.

• Advanced Settings

  You can modify the advanced parameters based on the resource pressure of serverless clusters.

  • Scan Interval: In Sensitive mode, serverless clusters with defined specifications respond to workloads in a faster manner, which reduces the observation window and execution period of the clusters. You can select Standard or Sensitive from the Scan Interval drop-down list.

  • Maximum CPU Resources for Elastic Upgrade: the CPU scale-up threshold for the cluster. Valid values: 40% to 100%.

  • Minimum CPU Resources for Elastic Upgrade: the CPU scale-down threshold for the cluster. Valid values: 10% to 70%.

  Note

  • The maximum CPU resources must be the same as or higher than the minimum CPU resources. The difference between the maximum CPU resources and the minimum CPU resources must be equal to or larger than 30 PCUs.

  • Serverless clusters in sensitive mode respond to load changes faster than clusters in standard mode and are suitable for businesses with instantaneous load fluctuations (such as instantaneous CPU usage spikes). Such clusters are frequently scaled based on load fluctuations.

Lifecycle Policy

Create a lifecycle policy to scale the cluster at specified time points monthly, weekly, or daily within a specified time range. It allows you to scale up or out the cluster during foreseeable peak hours (such as promotions and traffic spikes) to guarantee performance, and scale back the cluster during off-peak hours to release idle resources.

1. In the Configure Serverless-related Parameters dialog box, click + Add Lifecycle Policy. Set the following parameters:

   Parameter	Valid values
   Maximum Resources for Single Node	1 to 32.
   Minimum Resources for Single Node	1 to 32. The value of the Minimum Resources for Single Node parameter must be less than or equal to the value of the Maximum Resources for Single Node parameter.
   Maximum Read-only Nodes	0 to 15.
   Minimum Read-Only Nodes	0 to 15. The value of the Minimum Read-only Nodes parameter must be less than or equal to the value of the Maximum Read-only Nodes parameter.
   Read-only Column Store Nodes	0 to 15.
   Start/End Time	The validity period of the recurring policy.
   Policy Scheduling	The scheduling settings of the recurring policy. Valid values for the options of the Every parameter: Month: specifies the counting method of days and the days and time for execution. Positive indicates forward counting from the first day of a month. Last indicates backward counting from the last day of a month. Separate multiple days with commas (, ). Example: 1,3,5. Weekly: Specifies the day and time for execution. Daily: Specifies the time for execution.

   Note

   After a lifecycle policy is created, the system adjusts the serverless parameters of the cluster within the time range specified by Start/End Time based on the schedule specified by Policy Scheduling. Adjusted parameters are not automatically changed back. You can create another lifecycle policy to change back the parameters at a specified point in time. For more information, see Example.

2. (Optional) Query scheduled tasks. You can use one of the following methods to query scheduled tasks:

   Note

   After a recurring policy is created, scheduled tasks are automatically generated based on the policy.

   • After a recurring policy is created for a cluster, you can view scheduled tasks in the Pending Scheduled and Failed Tasks section on the Basic Information page of the cluster.

     

   • Choose Task Management > Scheduled Tasks on the PolarDB console to view scheduled tasks.

     

Example

If you want to scale up 5 PCUs at 9:30 and scale down 1 PCU at 22:00 every business day (Monday to Friday) from August 1 to September 30, you can configure the settings as shown in the following figures.