DRAForge
DRAForge is a Kubernetes Dynamic Resource Allocation (DRA) observability, simulation, and diagnostics platform for GPU and accelerator workloads.
It helps Kubernetes platform engineers model DRA resources, simulate virtual device pools, inspect ResourceClaims and ResourceSlices, and diagnose why dynamic resource allocations succeed or fail.
Features
- Kubernetes DRA diagnostics for ResourceClaims, ResourceSlices, drivers, and simulated device pools.
- Virtual accelerator simulation for GPU and device-style workloads.
- Explain engine for allocation troubleshooting and remediation.
- Terminal UI and React dashboard for resource visualization.
- Helm, Terraform, and DigitalOcean Kubernetes showcase assets.
Documentation
Showcase Demo Notes
When running the DRAForge showcase demo, the dashboard provides a real-time view of the cluster allocation state.
Demo Walkthrough Steps
- Interactive Graph: Navigate to the GRAPH tab to visualize the two worker nodes, resource slices, and claims.
- Cluster Health: Open the DOCTOR tab to see running status diagnostics (such as DRA API availability and version compatibility).
- Simulation Diagnostics: Go to the EXPLAIN tab and select any claim from the dropdown list to see the reason tree and suggested remediation steps.
- Relationship Scanning: Click on any node in the graph, and use the Relationships panel in the bottom-right corner to jump directly between pods, resource claims, and devices.
Repository
Keywords
Kubernetes DRA, Dynamic Resource Allocation, ResourceClaim, ResourceSlice, device plugin, GPU simulator, accelerator simulator, observability, diagnostics, Kubernetes troubleshooting, Go, React, TypeScript, Helm, Terraform, DigitalOcean Kubernetes.