Configuring Auto Scaling of vCPUs and Memory

Scenarios

RDS for PostgreSQL allows you to set limits on CPU usage. When the average CPU usage of your DB instance reaches the configured limits, the vCPU and memory specifications will be scaled up or down. Auto scaling of vCPUs and memory can handle changes in database load, helping you optimize resource costs and utilization.

Constraints

• Only primary/standby instances with the cloud SSD or extreme SSD storage type support auto scaling. If auto scaling is enabled for an instance, its storage type cannot be changed to any other type than cloud SSD or extreme SSD.
• To configure auto scaling, you must have related permissions. For details, see Permissions.
• To request the permissions required for auto scaling, submit a service ticket on the management console.
• Auto scaling of instance specifications will briefly interrupt services.
• Auto scaling of instance specifications cannot be triggered if the instance is abnormal, stopped, or frozen.
• Auto scaling of instance specifications cannot be enabled if the instance has any other ongoing tasks or scheduled changes in the specifications or storage type. If it has already been enabled, it still cannot be triggered under these conditions.
• Auto scaling of instance specifications cannot be triggered for any yearly/monthly instance that has unfinished orders or is in arrears.
• If your instance has read replicas, their specifications should not be too small, or the read replicas may have excessive latency and a heavy load. When the specifications of the primary instance are automatically scaled, the selected specifications must be no greater than those of the read replicas. To autoscale the specifications of the primary instance to be greater than those of the read replicas, submit a service ticket.
• There are temporary IP addresses used when auto scaling is performed, and they are not released immediately after the scaling is complete. There is a 12-hour delay by default. If there are not enough floating IP addresses available, the scaling will fail.
• During a scale-up, the system searches for a specification with more vCPUs and at least the same memory as the current amount. If the next larger matching specification is sold out or unavailable, the system continues to search for a higher one that matches the requirements. During a scale-down, to avoid out of memory (OOM) problems, if the next smaller specification is sold out or unavailable, the system will stop matching downwards. In addition, the CPU-to-memory ratio after the scale-down will remain the same as that of the original specification.
• If auto scaling of storage or specifications is required, you need to create an agency (agency name: RDSAccessProjectResource) and grant specific permissions to IAM users. For details, see Permissions. If the agency is canceled after the function is enabled, auto scaling will not work.

Billing

The billing for auto scaling is the same as that for changing an instance class. For details, see Billing.

Configuring Auto Scaling

1. Log in to the RDS console.
2. Click in the upper left corner and select a region.
3. On the Instances page, click the name of the target primary/standby instance to go to the Overview page.
4. Click Configure under Auto Scaling.
5. In the displayed dialog box, configure the required parameters.
   Figure 1 Setting auto scaling parameters
   

   Table 1 Parameter description

   Parameter	Description
   Auto Scale-up	You can enable or disable it as needed.
   Observation Period	When auto scale-up is enabled, the system periodically checks the vCPU usage. If the average vCPU usage exceeds the preset limit within the observation period, the system adapts the specifications to the read and write traffic. The system then enters a silent period. The minimum observation period is 5 minutes.
   Max. Specifications	The maximum specifications after the final auto scale-up. The system scales up specifications level by level. Each time the specifications are upgraded to a higher level, the system then enters a silent period.
   Trigger If Avg. vCPU Usage Reaches	The vCPU usage limit for triggering an auto scale-up. If the average vCPU usage exceeds this limit during the observation period, an auto scale-up is triggered.
   Auto Scale-down	You can enable or disable it as needed.
   Min. Specifications	The minimum specifications after the final auto scale-down. The system scales down specifications level by level. Each time the specifications are downgraded to a lower level, the system then enters a silent period.
   Trigger If Avg. vCPU Usage Drops To	The vCPU usage limit for triggering an auto scale-down. When the average vCPU usage in the observation period is less than this limit, an auto scale-down is triggered.
   Silent Period	The silent period is the minimum interval between two specification changes (triggered automatically or manually). During the silent period, RDS for PostgreSQL will not trigger auto scaling.

6. Click OK.

Checking the Change History

1. On the Instances page, click the name of the target primary/standby instance to go to the Overview page.
2. Click Change History under Auto Scaling.
3. In the displayed dialog box, check the change time, change type, status, original specifications, and new specifications of the historical tasks.