Auto Scaling:Target tracking scaling rules

Auto-created event-triggered tasks

When you create a target tracking scaling rule, Auto Scaling automatically creates two event-triggered tasks:

These tasks cannot be modified or deleted directly. To remove them, delete the target tracking scaling rule. You can view, disable, and enable these tasks.

Instance count calculation

Auto Scaling calculates the exact number of instances for each scaling activity based on metric history and target value:

• Scale-out: Rounds up. If 1.5 instances are needed, 2 instances are added.

• Scale-in: Rounds down. If 1.5 instances can be removed, only 1 instance is removed.

No scaling activity occurs if the metric data does not reach the threshold. If the calculated number of instances is less than one, alerts are continuously reported but no scale-in is triggered.

Supported metrics

Metrics used for target tracking must represent instance workload and scale proportionally with the number of instances in the scaling group.

ECS-type scaling groups

Elastic Container Instance-type scaling groups

Disable scale-in

By default, a target tracking scaling rule creates both scale-out and scale-in event-triggered tasks. Turn on Disable Scale-in to prevent the rule from removing instances.

When Disable Scale-in is on:

• Only a scale-out event-triggered task is created. No scale-in task is created.

• Use other methods to handle scale-in, such as a separate event-triggered task with a simple scaling rule.

When Disable Scale-in is off, both scale-out and scale-in tasks are created.

Configure Disable Scale-in through either method:

• Console: In the Create Scaling Rule dialog box, turn on the toggle next to Disable Scale-in.

• API: Set DisableScaleIn to true in the CreateScalingRule operation.

Configure instance warmup time

Instance Warmup Time is the period a newly added instance needs before it can serve production traffic. During warmup, Auto Scaling does not collect metric data from the instance. This prevents premature scaling triggered by incomplete metric data.

Set the warmup period based on the time required to deploy services, pass Server Load Balancer (SLB) health checks, and begin reporting stable metrics.

Behavior during warmup

• The instance belongs to the scaling group but does not contribute to CloudMonitor metrics.

• Instances being warmed up are not counted as the base for scale-out calculations.

• During the warmup period, Auto Scaling rejects requests to execute scaling rules in the scaling group.

• After the warmup period expires, the instance begins reporting data to CloudMonitor and is counted as part of the scaling group capacity.

Example: A scaling group contains 2 instances. A scale-out adds 5 new instances with a 300-second warmup period. Until the warmup expires, only the original 2 instances are used as the base for any subsequent scale-out calculation.

Scale-in protection during warmup

Auto Scaling sets an appropriate default cooldown time during scale-in to prevent removing instances that are still warming up due to data latency.

Create a target tracking scaling rule

Create a target tracking scaling rule through the Auto Scaling console or by calling the API:

• Console: Navigate to the scaling group, then create a scaling rule. For step-by-step instructions, see Manage scaling rules.

• API: Call the CreateScalingRule operation.

Limits

• Each metric can only be assigned to one target tracking scaling rule per scaling group.

• If the scaling group has a small number of instances, the metric value may differ significantly from the target value because each instance has a large impact on the aggregate metric.

Target tracking vs. simple scaling rules