The Autoscaling Rules page lets you create rules for when to scale out, in, up, or down based on memory usage.
- On the Application Console, select Auto Scaling from the Menu.The Autoscaling Rules page appears.
- On the Autoscaling Rules page, click Create Rule.The New Rule dialog appears.
- In the New Rule dialog, edit the following settings:
For example: Perform Scale Out to Maximum Cluster Size of 5 whenever Average of Memory Utilization is >= 20% for at least 2 consecutive periods of 5 minutes on All Instances and wait for 30 minutes of cool down period.
- Perform — Select Scale Up to increase memory per application instance, Scale Out to create additional application instances, Scale Down to decrease memory per application instance, Scale In to reduce number of application instances.
- GB Memory — Specify the maximum or minimum amount of memory per application instance that can exist as a result of a Scale Up or Scale Down rule.
- Cluster Size — Specify the maximum or minimum number of application instances that can exist as a result of a Scale Out or Scale In rule.
- Aggregation — Select Average, Maximum, or Minimum.
- Metric — Select Memory Utilization for percent memory usage or Memory for absolute memory usage in gigabytes.
- Threshold — Specify the value for the aggregated metric that triggers the rule.
- Consecutive Periods — Specify the number of consecutive periods for which the threshold condition must be met.
- Minutes — Specify the length of each period in minutes.
- Instances — Select All for all instances or select a specific instance.
- Cool Down — Specify the length of time in minutes during which the rule can’t be triggered again.
- Click Create.The new rule is listed on the Autoscaling Rules page. Events that activate the rule appear in the Alert History list.
- To return to the Application Console, select Open Application from the Menu.
After you create a rule, you can select Edit, Delete, or View Metrics from the Menu for the rule. View Metrics displays a list of metrics on which the rule is based.