Which metrics are essential to evaluate the state of your cloud infrastructure? Probably a lot. A dashboard allows you to keep an eye on all these metrics. For example, I like to monitor the following metrics for a typical 3-tier web application with the help of a CloudWatch dashboard:

• load balancer: client-side and server-side error rates
• load balancer: target response latency and number of requests
• compute: CPU and memory utilization
• database: query and I/O throughput

But how to create a customized CloudWatch Dashboard with CloudFormation? Doing so is possible with a simple CloudFormation template. However, I choose to use a custom resource to be more flexible when generating the dashboard. The following steps are needed to create a CloudWatch dashboard with a custom resource.

1. Create a CloudFormation template and add a Lambda-backed custom resource.
2. Write the code creating, updating, and deleting a CloudWatch dashboard.
3. Generate the JSON code describing the customized dashboard.

You will learn how to do so in more detail next.

Create a template and add a Lambda-backed custom resource

First of all, create a CloudFormation template. Add the identifiers of the resource you want to monitor to the parameters section as shown in the following code snippet.

---
Parameters:
  DashboardName:
    Description: 'The name for the dashboard.'
    Type: String
  AlbFullName:
    Description: 'The full-name of the ALB. (optional)'
    Type: String
    Default: ''
  RdsClusterName:
    Description: 'The RDS Aurora Cluster name. (optional)'
    Type: String
    Default: ''
# ...

Next, you need to add three resources to your template.

1. A custom resource Dashboard to manage the CloudWatch dashboard.
2. A Lambda function CustomResourceFunction executing your source code for the custom resource.
3. An IAM role CustomResourceRole assumed by the Lambda function to write logs as well as creating, updating and deleting CloudWatch dashboards.

# ...
Resources:
  Dashboard:
    Type: 'Custom::Dashboard'
    Version: '1.0'
    Properties:
      DashboardName: !Ref DashboardName
      AlbFullName: !Ref AlbFullName
      RdsClusterName: !Ref RdsClusterName
      ServiceToken: !GetAtt 'CustomResourceFunction.Outputs.Arn'
  CustomResourceFunction:
    Type: 'AWS::Lambda::Function'
    Properties:
      Code: './dashboard.js'
      Handler: 'dashboard.handler'
      MemorySize: 512
      Role: !GetAtt 'CustomResourceRole.Arn'
      Runtime: 'nodejs8.10'
      Timeout: 30
  CustomResourceRole:
    Type: 'AWS::IAM::Role'
    Properties:
      AssumeRolePolicyDocument:
        Version: '2012-10-17'
        Statement:
        - Effect: Allow
          Principal:
            Service: 'lambda.amazonaws.com'
          Action: 'sts:AssumeRole'
      Policies:
      - PolicyName: 'customresource'
        PolicyDocument:
          Statement:
          - Effect: Allow
            Action:
            - 'logs:CreateLogStream'
            - 'logs:PutLogEvents'
            Resource: !GetAtt 'LogGroup.Arn'
          - Effect: Allow
            Action:
            - 'cloudwatch:DeleteDashboards'
            - 'cloudwatch:PutDashboard'
            Resource: '*'

And now we’ll write some code.

Implement creating, updating, and deleting a CloudWatch dashboard

The following code snippet shows parts of dashboard.js including the handler which creates, updates, or deletes a CloudWatch dashboard. The built-in modules aws-sdk and cfn-response are used.

const AWS = require('aws-sdk');
const cloudwatch = new AWS.CloudWatch({apiVersion: '2010-08-01'});
const response = require('cfn-response');

exports.handler = (event, context, cb) => {
  console.log(`Executing function with event: ${JSON.stringify(event)}`);
  const error = (err) => {
    console.log('Error', err);
    response.send(event, context, response.FAILED);
  };
  if (event.RequestType === 'Delete') {
    cloudwatch.deleteDashboards({DashboardNames: [event.ResourceProperties.DashboardName]}, function(err) {
      if (err) {
        error(err);
      } else {
        response.send(event, context, response.SUCCESS, {}, event.ResourceProperties.DashboardName);
      }
    });
  } else if (event.RequestType === 'Create' || event.RequestType === 'Update') {
    cloudwatch.putDashboard({
      DashboardName: event.ResourceProperties.DashboardName,
      DashboardBody: JSON.stringify(generateDashboard(event))
    }, function(err) {
      if (err) {
        error(err);
      } else {
        console.log(`Created/Updated dashboard ${event.ResourceProperties.DashboardName}.`);
        response.send(event, context, response.SUCCESS, {}, event.ResourceProperties.DashboardName);
      }
    });
  } else {
    error(new Error(`unsupported request type: ${event.RequestType}`));
  }
};

There is only one step missing: you need to add widgets to the dashboard.

Generate the JSON code describing the customized dashboard

To do so, you need to generate a JSON file describing your dashboard in Dashboard Body Structure and Syntax . I recommend to create your dashboard by clicking through the AWS Management Console first. Generating the dashboard’s JSON with the help of the View/edit source action gives you a good starting point.

The following code snippet shows the generateDashboard function, which generates the JSON defining a CloudWatch dashboard with two widgets:

1. Widget ALB: total number of incoming requests as well as the target latency in 99 and 95 percentile
2. Widget RDS: maximum read and write IOPS as well as the serverless capacity of the database

function generateDashboard(event) {
  let widgets = [];
  if (event.ResourceProperties.AlbFullName) {
    widgets.push({
      type: 'metric',
      properties: {
        metrics: [
          ['AWS/ApplicationELB', 'TargetResponseTime', 'LoadBalancer', event.ResourceProperties.AlbFullName, {stat: 'p99', label: '99 PCTL'}],
          ['...', {'stat': 'p95', 'label': '95 PCTL' }],
          [{expression: 'SUM(METRICS(\"req\"))/60/5', label: 'Requests', id: 'reqs', yAxis: 'right'}],
          ['AWS/ApplicationELB', 'RequestCount', 'LoadBalancer', event.ResourceProperties.AlbFullName, {stat: 'Sum', id: 'req', visible: false }]
        ],
        view: 'timeSeries',
        stacked: true,
        region: event.ResourceProperties.Region,
        title: 'ALB Requests + Latency',
        period: 300
      }
    });
  }
  if (event.ResourceProperties.RdsClusterName) {
    widgets.push({
      type: 'metric',
      properties: {
         metrics: [
          ['AWS/RDS', 'VolumeReadIOPs', 'DBClusterIdentifier', event.ResourceProperties.RdsClusterName, {stat: 'Maximum'}],
          ['.', 'VolumeWriteIOPs', '.', '.', {'stat': 'Maximum'}],
          ['.', 'ServerlessDatabaseCapacity', '.', '.', {stat: 'Maximum', yAxis: 'right'}]
         ],
         view: 'timeSeries',
         region: event.ResourceProperties.Region,
         title: 'RDS IOPS + Capacity'
      }
    });
  }
  // Add additional widgets here ...

  let dashboard = {widgets: []};
  for (let w in widgets) {
    let widget = widgets[w];
    let x = (w % 2) * 12;
    let y = Math.floor(w/2);
    widget.x = x;
    widget.y = y;
    widget.width = 12;
    widget.height = 6;
    dashboard.widgets.push(widget);
  }

  console.log(`Generated dashboard: ${JSON.stringify(dashboard)}`);
  return dashboard;
}

That’s it. The following screenshot illustrates the results.

Looking for a more advanced example? Check out the source code of our CloudFormation modulecfn-modules/cloudwatch-dashboard.

Let’s knuckle down! 👩‍💻

🎉 Good news: our new book Rapid Docker on AWS is almost finished. 📚 We have handed over the last missing chapter to our editor today. Get your copy of the ebook and source code now and learn how to get your containerized web application up and running on AWS.

Published on 25 Sep 2019

Andreas Wittig

I’m the author of Amazon Web Services in Action. I work as a software engineer, and independent consultant focused on AWS and DevOps.

You can contact me via Email, Twitter, and LinkedIn.

Hire me

New book: Rapid Docker on AWS

A rapid way to get your web application up and running on AWS. Made for web developers and DevOps engineers who want to dockerize their web applications and run their containers on Amazon Web Services. Prior knowledge of Docker and AWS is not required.

Buy now