draforge

DRAForge

DRAForge is a provider-neutral Kubernetes Dynamic Resource Allocation (DRA) observability, simulation, and diagnostics platform for GPU and accelerator workloads.

It helps platform engineers model virtual device pools, inspect ResourceClaim and ResourceSlice state, and explain why dynamic resource allocations succeed or fail—without requiring physical accelerator hardware or a specific cloud provider.

Latest release · GitHub repository · Discussions · Security policy · Support

Start here

Goal	Guide
Run the verified local kind stack	Installation guide
Install or operate DRAForge on Kubernetes	Installation guide and operations guide
Understand components and data flow	Architecture
Check supported Kubernetes DRA behavior	Kubernetes DRA compatibility
Diagnose pending claims or runtime failures	Troubleshooting guide
Explore the dashboard, CLI, and API	Dashboard guide and public API/CLI surface

What DRAForge includes

• Evidence-based diagnostics for ResourceClaim, ResourceSlice, drivers, and simulated device pools.
• Virtual accelerator simulation for GPU and device-style workloads.
• Explain engine for selector, capacity, node-affinity, and allocation troubleshooting.
• Terminal UI and React dashboard for live resource visualization.
• Provider-neutral Helm deployment for compatible Kubernetes clusters.
• Optional Terraform and DigitalOcean Kubernetes showcase assets that are not required for normal installation, testing, or releases.

Documentation map

Users and operators

• Installation guide
• Operations guide
• Troubleshooting guide
• Dashboard guide
• Simulator scenarios
• Kubernetes DRA compatibility
• Optional DigitalOcean showcase

Contributors and maintainers

• Architecture
• Architecture decisions
• Local quality gate
• Install E2E matrix
• Release process
• Artifact gate
• Maintainer checklist
• Contributing guide
• Governance

Project scope

DRAForge components use Kubernetes APIs and Helm contracts rather than provider-specific services. They can run on compatible managed Kubernetes offerings, self-managed clusters, on-premises environments, and local test clusters. Provider-specific infrastructure and registry assets in this repository are optional showcases.

Keywords

Kubernetes DRA, Dynamic Resource Allocation, ResourceClaim, ResourceSlice, device plugin, GPU simulator, accelerator simulator, observability, diagnostics, Kubernetes troubleshooting, managed Kubernetes, on-premises Kubernetes, kind, Go, React, TypeScript, Helm, Terraform.

This site is open source. Improve this page.