SSM Automation Document ("Runbook") to start instances, run a Command Document, then stop instances that were started

SSM-Automation-StartStopInstances.yml

# Update this automation document:
# aws ssm update-document \
#     --name "StartStopInstances" \
#     --document-format YAML \
#     --content "file://./StartStopInstances-Automation.yml" \
#     --document-version '$LATEST'

# Execute this automation using IAM Service Role (https://gist.github.com/atheiman/021f1852997d0c42cd581d86aafc672f):
# aws ssm start-automation-execution \
#   --region us-gov-west-1 \
#   --document-name "StartStopInstances" \
#   --document-version '$LATEST' \
#   --parameters '{"AutomationAssumeRole":["arn:aws:iam::111111111111:role/AWS-SystemsManager-AutomationExecutionRole"]}'

---
description: >-
  Start EC2 instances by tag, run a Command Document, stop EC2 instances by tag.
schemaVersion: "0.3"
assumeRole: "{{ AutomationAssumeRole }}"
parameters:
  AutomationAssumeRole:
    type: String
    description: (Optional) The ARN of the role that allows Automation to perform the actions on your behalf.
    default: ""
outputs:
  - StartInstances.InstanceIds
  - StartInstances.StartingInstances
  - StopInstances.InstanceIds
  - StopInstances.StoppingInstances
mainSteps:
  - name: DescribeInstances
    maxAttempts: 3 # Retry 2 times if failure encountered
    action: aws:executeScript
    inputs:
      Runtime: python3.7
      Handler: handler
      Script: |
        import boto3

        def handler(event, context):
          print(event)
          print(context)
          res = {'StartInstances': [], 'StopInstances': []}
          ec2_resource = boto3.resource('ec2', region_name=context['global:REGION'])
          for i in ec2_resource.instances.all():
            # Find list of instances not opting out of Start
            if list(filter(lambda t: t['Key'].upper() == 'START' and t['Value'].upper() != 'FALSE', i.tags)):
              res['StartInstances'].append(i.instance_id)
              # Find same list for Stop, only stop the instance if we started it
              if list(filter(lambda t: t['Key'].upper() == 'STOP' and t['Value'].upper() != 'FALSE', i.tags)):
                res['StopInstances'].append(i.instance_id)
          return res
    outputs:
      - Name: StartInstances
        Selector: "$.Payload.StartInstances"
        Type: StringList
      - Name: StopInstances
        Selector: "$.Payload.StopInstances"
        Type: StringList

  - name: StartInstances
    maxAttempts: 3 # Retry 2 times if failure encountered (IncorrectInstanceState error)
    action: aws:executeAwsApi
    inputs:
      Service: ec2
      Api: StartInstances
      InstanceIds: "{{ DescribeInstances.StartInstances }}"
    outputs:
      - Name: StartingInstances
        Selector: "$.StartingInstances"
        Type: MapList
      - Name: InstanceIds
        Selector: "$..InstanceId"
        Type: StringList

  - name: WaitForInstancesStartup
    action: aws:sleep
    inputs:
      # Duration: PT10M # 10 minutes
      Duration: PT10S # 10 seconds

  # These steps should be updated with the actual actions to be taken against instances.
  # It is better to create a standalone SSM Command Document and Run it from this step, rather
  # than declaring the 'commands' inline as is shown below. In a full SSM Command Document, steps can
  # be designated to only only run on Windows or Linux instances using 'precondition' and
  # 'platformType'.
  - name: RunLinuxShellScript
    action: 'aws:runCommand'
    onFailure: step:StopInstances # Even if the script fails, be sure to stop instances
    inputs:
      DocumentName: AWS-RunShellScript # Only runs against Linux instances
      Targets:
        - Key: tag:Managed
          Values:
            - 'True'
      Parameters:
        commands:
          - echo "Running shell '${0}' as user '$(whoami)'"
          - 'echo "System info: $(uname -a)"'
  # Example for Windows instances
  # - name: RunPowerShellScript
  #   action: 'aws:runCommand'
  #   onFailure: step:StopInstances # Even if the script fails, be sure to stop instances
  #   inputs:
  #     DocumentName: AWS-RunPowerShellScript # Attempts to run against both Windows and Linux instances
  #     Targets:
  #       - Key: tag:Managed
  #         Values:
  #           - 'True'
  #     Parameters:
  #       commands:
  #         - Write-Output "Running PowerShell version '$(PSVersionTable.PSVersion)' as user '$(whoami)'"
  #         - >-
  #             Write-Output "Windows version info: '$(Get-WmiObject -Class Win32_OperatingSystem |
  #             Select-Object -Property Version,Caption)'"

  - name: StopInstances
    maxAttempts: 3 # Retry 2 times if failure encountered (IncorrectInstanceState error)
    action: aws:executeAwsApi
    inputs:
      Service: ec2
      Api: StopInstances
      InstanceIds: "{{ DescribeInstances.StopInstances }}"
    outputs:
      - Name: StoppingInstances
        Selector: "$.StoppingInstances"
        Type: MapList
      - Name: InstanceIds
        Selector: "$..InstanceId"
        Type: StringList