CloudWatch

• Monitoring tool for your AWS resources and applications.
• Can create alarms that watch the metrics and send notifications or automatically make changes to the resources when a threshold is breached.
• Metrics have timestamps.
• No data aggregation between regions.
• Metrics for every service on AWS.
• Metric: A variable to monitor (CPUUtilization, NetworkIn, …). You can find a complete list of all metrics here

Glossary

Name	Description	Side notes
Namespaces	A container for CloudWatch metrics.	Isolated from each other (To prevent aggregating different apps' statistics in one place) .
Metrics	A time-ordered set of data about the performance of your systems.	Region specific. By default gathered every 5 minutes. Expires , rather than being deletable. Delete data older than 15 month as new data comes in. Either specify timestamps or CloudWatch will create it for you. NO memory utilization and disk space usage metrics by default for EC2 instances.
Dimensions	A key/value pair that uniquely identifies a metric.	Whenever you add a unique dimension to one of your metrics, you are creating a new variation of that metric.
Statistics	Aggregated data over specified periods of time.	Specify a unit of measure are aggregated separately. Can specify a unit when you create a custom metric A period is the length of time associated with a specific CloudWatch statistic. Statistics are aggregated according to the period length that we specify.

Learn more

Some important metrics

Service	Metrics
EC2 instances	CPU Utilization. Status Checks. Network.
EBS Volumes	Disk read/write.
S3 Buckets	BucketSizeBytes. NumberOfObjects. AllRequests.
API usage	CloudWatch Lambda. S3.

But we can also create our own metrics.

CloudWatch alarms

• Goes off when a metric breaches a predefined threshold.
  • Can be based on percentage/max/min or other units.
  • Can configure how frequently check whether we breached the threshold or not.
  • Can be in different states:
    • OK.
    • INSUFFICIENT_DATA.
    • ALARM.
• Actions they can perform:
  • Send a notification.
  • Auto scaling.
  • Perform EC2 actions: stop/terminate/reboot/recover.
  • SNS notifications.
    1. You can easily create a new topic in SNS and subscribe to it with your email.
    2. Then in your CloudWatch you only need to create an alarm that publishes a message to that topic.
• Billing alarms can help us with budget management.

CloudWatch logs

• Collecting log files.
• Can collects from:
  • ECS.
  • Lambda.
  • Route 53.
  • CloudTrail.
  • CloudWatch log agents installed on <table style="display:inline"><tr><td>EC2 instances¹</td></tr><tr><td>on-premises servers</td></tr></table>.
  • Elastic Beanstalk.
• You can monitor logs in real-time.
• Configurable in terms of retention; for how long should I keep them.

Footnotes

1. For this to work we need proper IAM role. ↩