View a markdown version of this page

Deploy Deadline Cloud farms with the AWS CDK - Deadline Cloud

Deploy Deadline Cloud farms with the AWS CDK

The starter_farm AWS CDK app in the deadline-cloud-samples repository on the GitHub website deploys Deadline Cloud farms from reusable TypeScript constructs that wrap the CfnFarm, CfnQueue, and CfnFleet L1 resources from aws-cdk-lib/aws-deadline. For a comparison of the AWS CDK with CloudFormation and Terraform, see Manage Deadline Cloud infrastructure as code.

The app contains four example farms, each a separate AWS CDK stack:

  • SimpleFarm – A single Linux fleet on one queue. Deploy SimpleFarm when you are getting started.

  • StarterFarm – Adds a private conda channel and a queue that builds packages for it.

  • CudaFarm – A GPU fleet, with conda-forge for the CUDA toolchain.

  • MultiPlatformFarm – One queue reaching Linux, Windows, and GPU fleets.

Every stack creates a farm and an Amazon S3 bucket for job attachments. Each stack also includes at least one queue with a conda queue environment, at least one service-managed fleet, and one IAM role for each queue and each fleet.

To deploy, you need Node.js 18 or later, AWS credentials, an account and region bootstrapped for the AWS CDK, and a Deadline Cloud monitor. Then run:

cd cdk/farm_templates/starter_farm
npm ci
npx cdk deploy SimpleFarm