Dynamodb autoscaling using a combination of CloudFormation and AWS SDK

build.gradle

// snip: imports, applys etc.

// snip: assume 'deploy' task exists that deploys Serverless

task addScalingPolicy(dependsOn: deploy) {
    description 'CloudFormation cannot do this yet, so we do it via SDK'
    doLast {
        final asClient = AWSApplicationAutoScalingClientBuilder.defaultClient()
        final putScalingPolicy = { String tableName ->
            final ttspReadConfig = new TargetTrackingScalingPolicyConfiguration(
                    predefinedMetricSpecification: new PredefinedMetricSpecification(
                            predefinedMetricType: 'DynamoDBReadCapacityUtilization'),
                    targetValue: 70.0)
            final ttspWriteConfig = new TargetTrackingScalingPolicyConfiguration(
                    predefinedMetricSpecification: new PredefinedMetricSpecification(
                            predefinedMetricType: 'DynamoDBWriteCapacityUtilization'),
                    targetValue: 70.0)

            final scalingPolicyRequest = new PutScalingPolicyRequest(
                    serviceNamespace: 'dynamodb',
                    resourceId: "table/$tableName",
                    scalableDimension: 'dynamodb:table:WriteCapacityUnits',
                    policyName: "table/${tableName}_WriteScalingPolicy",
                    policyType: 'TargetTrackingScaling',
                    targetTrackingScalingPolicyConfiguration: ttspWriteConfig)

            asClient.putScalingPolicy(scalingPolicyRequest)

            scalingPolicyRequest.with {
                scalableDimension = 'dynamodb:table:ReadCapacityUnits'
                policyName = "table/${tableName}_ReadScalingPolicy"
                targetTrackingScalingPolicyConfiguration = ttspReadConfig
            }
            asClient.putScalingPolicy(scalingPolicyRequest)
        }

        putScalingPolicy 'myTable1'
        putScalingPolicy 'myTable2'
        // etc
    }
}

serverless.yml

# snip: various serverless-related provider settings etc

resources:
  Resources: # Essentially just CloudFormation
    myTable1DynamoTable:
      Type: AWS::DynamoDB::Table
      Properties:
        TableName: mytable1
        # insert AttributeDefinitions
        # insert KeySchema
        ProvisionedThroughput:
          ReadCapacityUnits: ${env:MYTABLE1_READCAP}
          WriteCapacityUnits: ${env:MYTABLE1_WRITECAP}
    myTable2DynamoTable:
      Type: AWS::DynamoDB::Table
      Properties:
        TableName: mytable2
        # insert AttributeDefinitions
        # insert KeySchema
        ProvisionedThroughput:
          ReadCapacityUnits: ${env:MYTABLE2_READCAP}
          WriteCapacityUnits: ${env:MYTABLE2_WRITECAP}
    myAutoscalingRole:
      Type: AWS::IAM::Role
      Properties:
        RoleName: DynamoDBAutoscaleRole
        AssumeRolePolicyDocument:
          Version: "2012-10-17"
          Statement:
          -
            Effect: Allow
            Principal:
              Service: "application-autoscaling.amazonaws.com"
            Action: "sts:AssumeRole"
        Policies:
          - PolicyName: "myAutoscalingPolicy"
            PolicyDocument:
              Version: "2012-10-17"
              Statement:
              - Effect: Allow
                Action:
                  - dynamodb:DescribeTable
                  - dynamodb:UpdateTable
                  - cloudwatch:PutMetricAlarm
                  - cloudwatch:DescribeAlarms
                  - cloudwatch:DeleteAlarms
                  - cloudwatch:GetMetricStatistics
                  - cloudwatch:SetAlarmState
                Resource: "*"
    myTable1ReadTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      DependsOn: myAutoscalingRole
      Properties:
        MaxCapacity: 10000
        MinCapacity: ${env:MYTABLE1_READCAP}
        ResourceId: table/mytable1
        RoleARN: {Fn::GetAtt: [myAutoscalingRole, Arn]}
        ScalableDimension: dynamodb:table:ReadCapacityUnits
        ServiceNamespace: dynamodb
    myTable1WriteTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      DependsOn: myAutoscalingRole
      Properties:
        MaxCapacity: 10000
        MinCapacity: ${env:MYTABLE1_WRITECAP}
        ResourceId: table/mytable1
        RoleARN: {Fn::GetAtt: [myAutoscalingRole, Arn]}
        ScalableDimension: dynamodb:table:WriteCapacityUnits
        ServiceNamespace: dynamodb
    myTable2ReadTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      DependsOn: myAutoscalingRole
      Properties:
        MaxCapacity: 10000
        MinCapacity: ${env:MYTABLE2_READCAP}
        ResourceId: table/mytable2
        RoleARN: {Fn::GetAtt: [myAutoscalingRole, Arn]}
        ScalableDimension: dynamodb:table:ReadCapacityUnits
        ServiceNamespace: dynamodb
    myTable2WriteTarget:
      Type: AWS::ApplicationAutoScaling::ScalableTarget
      DependsOn: myAutoscalingRole
      Properties:
        MaxCapacity: 10000
        MinCapacity: ${env:MYTABLE2_WRITECAP}
        ResourceId: table/mytable2
        RoleARN: {Fn::GetAtt: [myAutoscalingRole, Arn]}
        ScalableDimension: dynamodb:table:WriteCapacityUnits
        ServiceNamespace: dynamodb

This is an excellent gist on a little-documented CloudFormation feature. Note, in order to get the dynamodb scaling working all I needed were the correct CF resources shown here, and not the gradle SDK code.

Thanks for the example configurations! If you use serverless to create your CloudFormation Stack, have a look at my serverless-dynamodb-autoscaling plugin to configure Auto Scaling in your serverless.yml with a clean configuration:

custom:
  capacities:
    - name: custom-table  # DynamoDB table name
      read:
        minimum: 5        # Minimum read capacity
        maximum: 1000     # Maximum read capacity
        usage: 0.75       # Targeted usage percentage
      write:
        minimum: 40       # Minimum write capacity
        maximum: 200      # Maximum write capacity
        usage: 0.5        # Targeted usage percentage