Building CloudWatch dashboards

1. Building CloudWatch dashboards

Welcome back! In this video, we'll explore CloudWatch dashboard options.

2. Automatic dashboards

Automatic dashboards are pre-built, service-specific views CloudWatch generates as soon as you use a service. Zero configuration, free, and updating in real-time as your resources change. They come in a few types: a cross-service overview, per-service, and per-resource. They're great for quick health checks and discovering available metrics, but you can't customize them, add custom metrics, or share them externally. For that, you need custom dashboards.

3. Custom dashboards

Custom dashboards are collections of widgets that visualize metrics and alarms on a grid, with the column, row, metric, and widget limits shown here. You can create a few at no extra cost, but beyond that you're charged per dashboard. Create them interactively via the Console, or programmatically with the CLI, CloudFormation, or Terraform, so dashboards can be managed as code and deployed alongside your infrastructure.

4. When to use automatic vs. custom

Automatic dashboards cover a limited subset of AWS services, including EC2, Application ELB, Lambda, and DynamoDB. They're great for quick checks, but to dive deeper, bring multiple services together, and share insight with colleagues, you'll need custom dashboards.

5. Custom dashboard widgets

Custom dashboards offer eight widget types. Four are for showing data in graphs: stacked area charts, bar charts, line graphs, and pie charts. Each helps identify correlations, aggregate values, and trends over time. Stacked area charts are especially good when multiple metrics group up, like showing how each volume contributes to total storage.

6. Custom dashboard widgets

Number and gauge widgets are great for KPIs or a point-in-time value. With a gauge, we can set color-coded thresholds for low, medium, and high usage, for example showing a website's response time and whether it's fine or taking too long.

7. Custom dashboard widgets

Finally, text and log widgets add context alongside numeric data. Log widgets are typically paired with aggregates showing error, warning, or info message counts, while text widgets add specific knowledge and documentation.

8. Dashboard design principles

Five key principles guide good dashboards: group related metrics with the most important at the top; name dashboards by environment and purpose; use consistent widget widths on the 24-column grid; set relative time ranges with auto-refresh; and keep dashboards focused on actionable metrics. You can find more detail in our DataCamp tutorial on effective dashboard design.

9. Detecting performance issues

Build dashboards around your key performance metrics and the issues most likely to affect them. Four common patterns to watch for: CPU contention, high CPU dragging down response times despite normal traffic. External call latency, telling slow downstream dependencies apart from internal application issues. Intermittent errors, tracking sporadic spikes over time to spot seasonality and correlations. Memory leaks, gradual consumption that creeps up unnoticed until it suddenly hits performance.

10. Create a custom dashboard

Once we understand the audience and the insight they need, we can build out our dashboards. It's an iterative process: identify the metric or information, select the right widget, then set placement and size. I generally create all the widgets I need before sizing and arranging them, to avoid reworking large layouts. Once it's done, we share the dashboard with those who need it.

11. Sharing dashboards with others

Custom dashboards have three sharing methods: email-based access for up to five password-protected recipients, a public read-only link, or Single Sign-On for organization-wide access. When sharing, remember the principle of least privilege, don't share sensitive metrics publicly, and use text widgets to add context.

12. Let's practice!

Now let's put the information we have learned into practice.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.