For those who aren't familiar with Auto Scaling, it is an automation service Amazon provides as part of their cloud offering. It provides features to manage a running pool of servers, including the capability to replace failed instances and automatically grow and shrink the size of the pool. For a more thorough description, please see the Amazon Documentation.

• Auto Scaling Getting Started Guide
• Auto Scaling Developer Guide
• Auto Scaling API Reference
• Amazon CloudWatch Getting Started Guide
• Amazon CloudWatch Developer Guide
• Amazon CloudWatch API Reference

• Set min, desired and max to be the same for the group
  • They should also be set to a value greater than the capacity you think you need
• Disable scaling actions
  • Command:

as-suspend-processes MyAutoScalingGroup
as-resume-processes MyAutoScalingGroup

• AWS Quick Reference
• AWS Detailed Doc

See Amazon's documentation about the default metrics publish to CloudWatch

Requests-per-second (RPS) metics can be an important way to measure system performance. RPS is usually highly correlated with availability metrics: tomcat threads, apache workers, etc.

Time is also an important metric to monitor. When possible, avoid averages, in favor of percentiles. Specifically, 95th and 99th percentile.

To publish metrics directly to CloudWatch, refer to the documentation for the Amazon CloudWatch CLI or programmatic interfaces.

Command line example:

mon-put-data -namespace "NFLX/TEST" -metric-name "Foo" -value 123

Servo is an application monitoring library that allows you to register any fields in your Java code as custom metrics published to CloudWatch.

See the Servo documentation.

Refer to the wiki side bar for more documentation about auto scaling.