Example project demonstrating an AWS Batch infrastructure for processing CSV files from S3.

Builders often turn to AWS Lambda as an S3 trigger to process files as they are uploaded. While this may work for some use cases, Lambdas have the following limitations that may not fit this use case:

• 5 minute maximum
• Limited retries
• Low memory limit
• Limited instance-based storage

AWS Batch is a container-based solution that alleviates some of these concerns. EC2 spot can be used to lower the cost of the solution.

• AWS CodeCommit repository for the code repository
• AWS EC2 Container Registry Docker image registry
• AWS CodeBuild will build the Docker image and push to the ECR repository
• AWS Batch Compute Cluster will use a spot fleet at 25% of the on-demand price
• AWS Batch Job will run a python script to process the CSV in a streaming fashion

Terraform is used to describe the infrastructure.

• Create a file for the terraform variables (deploy/terraform.tfvars)

region = "us-west-2"
vpc_id = "vpc-XXXXXXX"
security_group_ids = [
    "sg-XXXXXXXX"
]

• Use Terraform to setup the infrastructure

cd deploy
terraform apply

• Push the code to your newly created CodeCommit repository

git remote add codecommit <SSH_URL>
git push codecommit master

• Submit a build

aws codebuild start-build --project-name aws_batch_csv_processor

• Once the build completes, a Batch job can be submitted

aws batch submit-job --job-name "first_run" --job-queue "batch_job_queue" --job-definition batch_csv_processor:1

• Once the job is complete, the logs will show up in CloudWatch logs

The CSV processor in the repository is quite simple (it just counts the columns and rows). But you can customize the script to do whatever you need to with the CSV!

The code is located in aws_batch_csv_processor/__main__.py. Most of the code is dedicated to reading the CSV file from S3 by streaming the contents.

The parameters to the batch job are the S3 bucket and key for the file:

• bucket is the S3 bucket parameter
• path is the path to the S3 object to read

By default, the job will read the IRS 990 public dataset for 2017.