AWS CloudWatch Basics: Metrics, Logs and Alarms for Your Resources

Last time I looked at who changed things with CloudTrail. The next question is different: what is happening inside my resources right now? That is Amazon CloudWatch.

What CloudWatch does

• Metrics: numbers over time, such as EC2 CPU utilization or S3 request counts.
• Logs: log data from your applications and services, stored in log groups.
• Alarms: watch a metric and do something when it crosses a threshold.
• Dashboards: put the graphs you care about on one page.

CloudTrail vs CloudWatch

How an alarm works

• An alarm watches one metric over a period (for example, 5 minutes).
• Its state is OK, ALARM or INSUFFICIENT_DATA.
• When the state changes, it can send a message to an SNS topic, which can email you.
• The threshold and the number of periods to evaluate decide how sensitive it is.

Good to know

• EC2 sends basic metrics every 5 minutes by default, and detailed monitoring gives 1-minute data (extra cost).
• EC2 does not report memory or disk usage by default. You need the CloudWatch agent for those.
• Logs stay forever unless you set a retention period, so set one.

Hands-on: try it in your own account

1. Open CloudWatch and browse All metrics for EC2 or S3.
2. Choose Alarms, then Create alarm.
3. Pick a metric, such as EC2 CPUUtilization for one instance.
4. Set a threshold, for example greater than 70% for 2 periods.
5. Create or choose an SNS topic with your email, and confirm the subscription email.
6. Save the alarm and check that it shows OK or INSUFFICIENT_DATA at first.

Common mistakes

• Creating alarms with no action, so nobody is told.
• Forgetting to confirm the SNS email subscription.
• Expecting memory metrics without the CloudWatch agent.
• Leaving log groups on never-expire retention.

Key takeaways

• CloudTrail tells you who changed it, CloudWatch tells you how it is doing.
• Metrics feed alarms, alarms feed SNS, SNS reaches you.
• Install the agent for memory and disk.
• Set log retention.

Tomorrow I'll look at how to use tags to keep resources organized and costs traceable.

Part of #120DaysOfAWS.