How to Create a CloudWatch Dashboard for Multiple EC2 Instances

By Ravindra Bagale, trainer · LinkedIn · Last updated: 21 October 2026

1. A story: the war room on sale night

On sale night the team sits in one room with a big TV. The manager asks every few minutes: "Are we fine?" Nobody wants to open 10 console pages. The TV shows one dashboard: CPU of all web servers, requests per minute, 5xx errors, the alarm lights and the latest error logs. One look answers the question.

2. Dashboards and widgets

Why do we need this?

Graphs in different places are slow to compare. When CPU jumps, you want the request count and the error logs right next to it.

What is it?

A dashboard is a page you design. A widget is one box on it.

3. Lab: build an app dashboard in the console

You need 2-3 running instances (tag Role=web), the alarms from the earlier chapters and the /app/nginx/access log group. With fewer pieces, build only the widgets you can.

Part A: create the dashboard

1. Open CloudWatch → Dashboards → Create dashboard.
2. Name: shop-prod. Why a clear name: people open it under stress; the name must say which app and which environment.
3. Click Create dashboard. The "Add widget" window opens.

Part B: CPU of all web servers on one graph

4. Choose Line → Metrics → Next.
5. Click EC2 → Per-Instance Metrics. In the search box type CPUUtilization.
6. Tick CPUUtilization for each web instance. Each becomes one coloured line.
7. Open Graphed metrics. Set Statistic Average, Period 1 minute if detailed monitoring is on, else 5 minutes.
8. Change each Label from the ID to web1, web2, web3. Why: IDs mean nothing to people at 2 AM.
9. Options tab: add a horizontal annotation at 80 labelled alarm line. Why: you see how close you are to the alarm.
10. Click Create widget. Title it Web CPU % (click the title to edit).

Faster for many servers: instead of ticking each one, use a search expression in the Source tab. New Auto Scaling servers appear by themselves:

SEARCH('{AWS/EC2,InstanceId} MetricName="CPUUtilization"', 'Average', 300)

Part C: number, metric math and memory

11. + Add widget → Number → Metrics. Pick the ALB metric RequestCount (ApplicationELB → Per AppELB Metrics), statistic Sum, 1 minute. Title Requests / min.
12. Add a line widget with HTTPCode_Target_5XX_Count (m1, Sum) and RequestCount (m2, Sum). Click Add math → Start with empty expression: 100*m1/m2, label 5xx %. Untick m1 and m2. Why: "12 errors" means little; "0.4% of requests failed" is clear.
13. Add a Gauge on CWAgent → InstanceId → mem_used_percent for web1, range 0-100. Title web1 memory %.

Part D: alarms and logs on the same screen

14. + Add widget → Alarm status. Tick web1-cpu-high, web1-system-recover, web1-instance-reboot, web1-disk-queue. Create.
15. + Add widget → Logs table. Log group /app/nginx/access. Query:

fields @timestamp, @logStream, @message
| filter @message like / 50[0-9] /
| sort @timestamp desc
| limit 20

16. Click Create widget. Title Latest 5xx lines.
17. + Add widget → Text: **On call:** Ops team · Runbook: link to your wiki. Why: the person who sees red knows what to do next.
18. Drag widgets so the most important ones (alarm status, 5xx %) are at the top. Click Save. Why save: unsaved changes are lost when you leave the page.
19. Top right: time range 3h, refresh 1 minute, and turn on Auto refresh for the TV.
20. Verify: run stress-ng --cpu 0 --timeout 300s on web1; its line climbs while the others stay flat.

4. The same dashboard with the CLI

Why do we need this?

A dashboard made by clicking exists in one account only. As JSON you can keep it in Git, review it and create it again in a new account or region in seconds.

How to do it (step by step)

1. Open your dashboard → Actions → View/edit source to see its JSON. Save a shorter version as dash.json:

{
  "widgets": [
    { "type": "metric", "x": 0, "y": 0, "width": 12, "height": 6,
      "properties": { "title": "Web CPU %", "region": "ap-south-1", "view": "timeSeries",
        "stat": "Average", "period": 300,
        "metrics": [
          ["AWS/EC2", "CPUUtilization", "InstanceId", "i-0aaa1111111111111", {"label": "web1"}],
          ["...", "i-0bbb2222222222222", {"label": "web2"}]
        ],
        "annotations": { "horizontal": [ { "value": 80, "label": "alarm line" } ] } } },
    { "type": "log", "x": 12, "y": 0, "width": 12, "height": 6,
      "properties": { "title": "Latest 5xx lines", "region": "ap-south-1", "view": "table",
        "query": "SOURCE '/app/nginx/access' | fields @timestamp, @logStream, @message | filter @message like / 50[0-9] / | sort @timestamp desc | limit 20" } },
    { "type": "text", "x": 0, "y": 6, "width": 24, "height": 2,
      "properties": { "markdown": "**On call:** Ops team" } }
  ]
}

2. Create or update it:

aws cloudwatch put-dashboard --dashboard-name shop-prod-cli --dashboard-body file://dash.json
# expected: {"DashboardValidationMessages": []}
aws cloudwatch list-dashboards --query 'DashboardEntries[].DashboardName'
# expected: ["shop-prod", "shop-prod-cli"]

"..." in the metrics array means "same as the line above", so you only write what changes (the instance ID).

5. Sharing a dashboard

Why do we need this?

A manager or a client wants to see the dashboard but should not get AWS console access. Giving console access to see graphs is too much power.

What is it?

How to do it (step by step)

1. Open the dashboard → Actions → Share dashboard.
2. Choose Share your dashboard and require a username and password. Enter the email addresses.
3. Click Prepare for sharing and confirm. AWS creates the needed role and an Amazon Cognito user pool behind the scenes.
4. Verify: each person gets an email with a link and sets a password; they see the dashboard read-only, with no console menu.
5. To stop sharing: Actions → Share dashboard → Stop sharing.

Cost: the free tier covers 3 dashboards with up to 50 metrics each per month. After that each dashboard is billed per month (about USD 3 at the time of writing; check the CloudWatch pricing page). Logs widgets also bill for the data each query scans.

6. Dashboard ideas per stack

7. Cleanup

1. Stop sharing if you shared it.
2. CloudWatch → Dashboards → tick shop-prod and shop-prod-cli → Delete.
3. CLI: aws cloudwatch delete-dashboards --dashboard-names shop-prod shop-prod-cli.
4. Delete the alarms, log groups and instances from the earlier labs if you no longer need them.

8. Interview questions

Revision, comparison and interview cards

Revise this lesson in a few minutes. Save the cards on your phone or download them as PNG files. The QR code on each card opens this lesson.

More cards: Ravi Sir's Cards.