Docs / Requirements

Requirements

Prepare these before creating the AWS stack.

RequirementNotes
AWS account and regionUse one primary region for the control plane, database, object store, and workers.
AWS credentialsThe deploying principal needs permission to create VPC, ECS, RDS, ElastiCache, S3, Secrets Manager, IAM, ALB, CloudFront, Auto Scaling, and EC2 resources.
OpenTofu or TerraformThe AWS profiles are OpenTofu-compatible. Use the infra shell if you want the repo-pinned toolchain.
AWS CLI and jqNeeded for reading outputs, writing secret values, and running the migration task.
GitHub OAuth appRequired for browser login.
Helmr release versionAWS examples read control image and worker AMI metadata from the release artifact manifest.
Public URLUse HTTPS for customer environments. Quickstart can use the generated CloudFront URL; production usually uses your own domain and ACM certificate.

Workers have additional requirements because they run Firecracker guests:

  • EC2 instance type with KVM support.
  • Worker AMI that includes helmr-worker, Firecracker, jailer, CNI plugins, tc-redirect-tap, BuildKit, AWS CLI v2, curl, kernel, initramfs, and rootfs artifacts.
  • Root EBS volume sized for filesystem-first build/cache/runtime data; worker_disk_mib can cap advertised capacity when needed.
  • Outbound access to the control plane, S3, ECR, AWS APIs, and any external services your tasks call.
  • SSM access for maintenance. Do not expose SSH by default.

You can deploy the control plane first and add workers later.

Schedules require both database and Redis/Valkey availability. The database stores schedule definitions, instances, and exact next fire times; Redis/Valkey stores replaceable next-fire entries and leases used by the dispatcher.