Having lambada pass off a codepipline task to an EC2 instance

codepipeline_handoff.py

import boto3
import os


def lambda_handler(event, context):
    job_id = event['CodePipeline.job']['id']
    s3_info = (event['CodePipeline.job']['data']
                    ['inputArtifacts'][0]['location']['s3Location'])

    client = boto3.client('ssm')
    client.put_parameter(
        Name='/codepipeline/jobid',
        Value=job_id,
        Type='String',
        Overwrite=True
    )
    client.put_parameter(
        Name='/codepipeline/artifact/zip',
        Value=s3_info['objectKey'],
        Type='String',
        Overwrite=True
    )
    client.put_parameter(
        Name='/codepipeline/artifact/bucket',
        Value=s3_info['bucketName'],
        Type='String',
        Overwrite=True
    )

    client = boto3.client('autoscaling')
    client.update_auto_scaling_group(
        AutoScalingGroupName=os.getenv('ASG_NAME'),
        MinSize=1,
        MaxSize=1
    )

Some experimentation with the thought of EC2 handoff with Lambda. Essentially, a Lambda invoked by CodePipeline passes off the job id, artifact zip, and artifact bucket name to EC2 parameter store, then modifies an ASG to have one instance launched.

The reason I use ASG here is that it automatically handles which AZ to put the instance in, and can also handle spinning up another instance should an AZ failure occur. Since the basic data is in parameter store I can use it to pickup where I left off when the new, healthy instance comes back up.