CloudWatch Alarms

• Alarms are used to trigger notifications for any metric.

• Alarms cab be raised for various options such as sampling, max, min etc.

• Alarm States

  • OK

  • INSUFFICIENT_DATA

  • ALARM

• Period

  • Length of time in seconds to evaluate the metric.

  • High resolution cutom metrics can be collected for periods like 10 secs, 30 secs or multiple of 60 secs.

• Targets

  • Alarms have a targets such as stop, restart or recover an EC2 instance.

  • Trigger Auto Scaling Action.

  • Send notifications to SNS

Composite Alarms

• CloudWatch Alarms are on a single metrics

• Composite Alarms are monitoring the states of multiple other alarms, basically AND and OR conditions.

• This helps creating complex composite alarms, so as to reduce alarm noise.

EC2 Instance Recovery

• ClouWatch Alarms can be created on following EC2 instance status Check,

  • Instance status -> Check EC2 VM

  • System status -> Check underlying hardware

  • Attached EBS status -> Check attached EBS volumes

• On breach of these alarms, one can start EC2 Instance Recovery, which moves an EC2 instance from one host to another.

• On recovery of EC2 instance, we get same private, public, elastic IP, metadata, Placement group as with previous host.