Terraform Provider for RunPod

A Terraform provider for managing RunPod GPU cloud resources.

Features

• Pod Management: Create, update, and delete GPU pods
• GPU Type Discovery: Query available GPU types and their specifications
• GPU Type Selection: Specify the GPU type for your pod

Requirements

• Terraform >= 1.0
• Go >= 1.24 (for building from source)
• A RunPod account and API key

Installation

From Source

git clone https://github.com/nilenso/terraform-provider-runpod.git
cd terraform-provider-runpod
make install

From Terraform Registry

Once published to the Terraform Registry, add to your required_providers:

terraform {
  required_providers {
    runpod = {
      source  = "nilenso/runpod"
      version = "~> 0.1"
    }
  }
}

Configuration

Provider Configuration

provider "runpod" {
  api_key = "your-api-key"  # Or use RUNPOD_API_KEY environment variable
}

Environment Variables

Variable	Description
RUNPOD_API_KEY	Your RunPod API key

Usage

Query Available GPU Types

# List all available GPU types
data "runpod_gpu_types" "all" {
}

# Get a specific GPU type
data "runpod_gpu_types" "a4000" {
  filter {
    id = "NVIDIA RTX A4000"
  }
}

output "gpu_types" {
  value = data.runpod_gpu_types.all.gpu_types
}

Create a GPU Pod

resource "runpod_pod" "example" {
  name       = "my-gpu-pod"
  image_name = "runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04"
  
  gpu_type_id        = "NVIDIA RTX A4000"
  gpu_count          = 1
  volume_in_gb       = 40
  container_disk_in_gb = 20
  
  # Optional settings
  cloud_type        = "ALL"  # ALL, SECURE, or COMMUNITY
  ports             = "8888/http,22/tcp"
  volume_mount_path = "/workspace"
  
  # Environment variables
  env = {
    JUPYTER_PASSWORD = "mysecretpassword"
  }
}

output "pod_id" {
  value = runpod_pod.example.id
}

Resources

runpod_pod

Manages a RunPod GPU pod.

Arguments

Attribute	Type	Required	Description
name	string	Yes	The name of the pod
image_name	string	Yes	Docker image to use
gpu_type_id	string	Yes	GPU type ID (e.g., "NVIDIA RTX A4000")
gpu_count	number	No	Number of GPUs (default: 1)
volume_in_gb	number	No	Persistent volume size in GB (default: 0)
container_disk_in_gb	number	No	Container disk size in GB (default: 20)
cloud_type	string	No	Cloud type: ALL, SECURE, COMMUNITY (default: ALL)
ports	string	No	Ports to expose (e.g., "8888/http,22/tcp")
volume_mount_path	string	No	Volume mount path (default: /workspace)
docker_args	string	No	Docker arguments
env	map(string)	No	Environment variables
min_vcpu_count	number	No	Minimum vCPUs required
min_memory_in_gb	number	No	Minimum memory in GB
network_volume_id	string	No	Network volume to attach
template_id	string	No	Template to use
data_center_id	string	No	Specific data center
support_public_ip	bool	No	Support public IP (default: true)
start_ssh	bool	No	Start SSH service (default: true)

Attributes (Read-Only)

Attribute	Description
id	The pod's unique identifier
machine_id	The machine ID the pod is running on
pod_host_id	The host ID of the pod

Import

Pods can be imported using their ID:

terraform import runpod_pod.example <pod-id>

Data Sources

runpod_gpu_types

Fetches available GPU types from RunPod.

Arguments

Attribute	Type	Required	Description
filter.id	string	No	Filter by GPU type ID

Attributes

Attribute	Description
gpu_types	List of GPU types
gpu_types[].id	GPU type ID (e.g., "NVIDIA RTX A6000")
gpu_types[].display_name	Display name
gpu_types[].memory_in_gb	GPU memory in GB
gpu_types[].secure_cloud	Available on secure cloud
gpu_types[].community_cloud	Available on community cloud

Development

Building

make build

Testing

# Unit tests
make test

# Acceptance tests (requires RUNPOD_API_KEY)
export RUNPOD_API_KEY="your-api-key"
make testacc

# Run a specific acceptance test
make testacc-one TEST=TestAccPodResource_lifecycle

Installing Locally

make install

Formatting and Linting

make fmt
make lint

License

MIT License - see LICENSE for details.