CloudWatch Metrics and Alarms

Welcome! In this lesson, we'll dive into AWS CloudWatch, a powerful monitoring service. CloudWatch helps you keep an eye on your AWS resources and applications.

We'll focus on two core features: Metrics and Alarms. These are essential for maintaining the health and performance of your cloud infrastructure.

At its heart, CloudWatch collects metrics. A metric is a time-ordered set of data points that represent a variable you want to monitor.

• Think of metrics as data points for performance, usage, or operational health.
• Examples include CPU utilization of an EC2 instance, network traffic, or disk I/O.
• AWS services automatically publish many metrics for you.

CloudWatch offers two main types of metrics:

• Standard Metrics: These are automatically collected by AWS services (like EC2, RDS, Lambda) for their resources. You get them out-of-the-box.
• Custom Metrics: You can define and publish your own application-specific metrics. This is useful for monitoring things like custom application latency, user logins, or specific business transactions.

You can easily view and graph metrics using the AWS Management Console.

Navigate to the CloudWatch service. From there, you can browse metrics by AWS service (e.g., EC2, Lambda) and then by resource ID. This allows you to visualize trends and understand resource behavior over time.

Metrics are organized using dimensions. A dimension is a name/value pair that uniquely identifies a metric.

• For example, an EC2 instance's CPU Utilization metric might have a dimension like InstanceId: i-1234567890abcdef0.
• Dimensions help you filter and segment your metric data, making it easier to find specific insights.
• You can add up to 10 dimensions to a custom metric.

While metrics show you what's happening, CloudWatch Alarms tell you when something needs attention. An alarm monitors a single metric over a specified period.

If the metric breaches a defined threshold, the alarm changes state and can perform actions, like sending a notification or initiating an Auto Scaling action.

CloudWatch Alarms have three main states:

• OK: The metric is within the defined threshold. Everything is normal.
• ALARM: The metric has breached the defined threshold. This indicates a potential issue.
• INSUFFICIENT_DATA: There isn't enough data to determine the alarm's state (e.g., new metric, data collection stopped).

Let's consider a practical example: creating an alarm for high EC2 CPU utilization.

You would select the CPUUtilization metric for a specific EC2 instance, set a threshold (e.g., > 80%), and define the evaluation period (e.g., 5 minutes for 3 consecutive periods). If the CPU stays above 80% for 15 minutes, the alarm triggers.

When an alarm changes state, it can trigger various actions. A common action is sending a notification using Amazon SNS (Simple Notification Service).

You can configure an SNS topic to receive alarm notifications, which can then deliver messages via email, SMS, or even trigger other AWS services like Lambda functions. This ensures you're alerted to issues promptly.

Which of the following are valid actions that a CloudWatch Alarm can take when it transitions to the ALARM state?

Great job! You've learned the fundamentals of CloudWatch metrics and alarms.

• Metrics are time-ordered data points for monitoring.
• Dimensions help organize and filter metrics.
• Alarms monitor metrics against thresholds.
• Alarms have OK, ALARM, and INSUFFICIENT_DATA states.
• You can configure actions (like SNS notifications) for alarm state changes.

Understanding these concepts is key to proactive monitoring and maintaining healthy AWS environments.