CI/CD Environment
Self-Hosted GitHub Actions Runners on Dedicated GPUs
Deploy self-hosted GitHub Actions runners that never queue, have full GPU access, and cost up to 70% less than GitHub's hosted runners. Pre-configured Ubuntu 24.04 VMs with Docker, ready in minutes.
TL;DR
- →Zero queue time — your runners are always warm and waiting for jobs
- →GPU access in CI — run ML training, model tests, and CUDA builds directly in your pipeline
- →70% cheaper — dedicated VMs from $0.29/hr vs GitHub's $0.48/min for GPU runners
- →Auto-failover — if a node goes down, runners migrate automatically
What Is the GitHub Runners Environment?
The GitHub Runners environment is a pre-built VM template that deploys self-hosted GitHub Actions runners on VectorLay's infrastructure. Each VM comes with Ubuntu 24.04, Docker, and the GitHub Actions runner agent pre-installed. On boot, the runner automatically registers with your GitHub repository or organization and starts picking up jobs.
You configure a few parameters — your repo URL, a personal access token, and how many replicas you want — and VectorLay handles the rest. No SSH-ing into servers, no installing packages, no managing systemd services. The runners are production-ready from the moment they launch.
Why Self-Hosted Runners?
GitHub's hosted runners work fine for basic workflows, but they fall short when your CI/CD pipeline needs more than a generic Ubuntu container. Here's why teams switch to self-hosted runners on VectorLay:
What's Included
Every GitHub Runners environment VM boots with:
Ubuntu 24.04 LTS
The same base OS as GitHub's hosted runners. Full compatibility with existing workflows — no changes needed.
Docker Engine
Pre-installed and configured. Run containerized steps, build Docker images, and push to registries out of the box.
GitHub Actions Runner
Auto-registers with your repo or org on boot. Deregisters cleanly on shutdown. No manual token management.
NVIDIA Drivers + CUDA
GPU nodes include pre-installed NVIDIA drivers and CUDA toolkit. Run GPU workloads in CI without any driver setup.
How It Works
Select the Environment
Go to Environments in the VectorLay dashboard and select GitHub Runners.
Configure Your Runner
Choose repo-scoped or org-scoped, paste your repository URL and a GitHub Personal Access Token. Optionally add custom labels for workflow routing.
Set Replicas
Choose 1–10 runner replicas for parallel job execution. Each replica is an independent VM that can pick up jobs simultaneously.
Deploy
Hit deploy. The VM boots, registers with GitHub, and starts accepting jobs. You'll see it appear as an active runner in your GitHub Settings within seconds.
Pricing: VectorLay vs. GitHub Hosted Runners
GitHub charges per minute. VectorLay charges per hour for a dedicated VM. For anything beyond light usage, self-hosted runners on VectorLay are dramatically cheaper.
| Provider | Runner Type | Cost | GPU? | Queue? |
|---|---|---|---|---|
| VectorLay | Dedicated VM (RTX 3090) | $0.29/hr | Yes | None |
| VectorLay | Dedicated VM (RTX 4090) | $0.49/hr | Yes | None |
| GitHub | Linux (2-core) | $0.008/min | No | Variable |
| GitHub | Linux (16-core) | $0.064/min | No | Variable |
| GitHub | GPU (4-core + T4) | $0.07/min | T4 only | Variable |
For a team running 8 hours of CI per day, VectorLay costs roughly $70/month for an RTX 3090 runner — compared to $230+/month for GitHub's hosted 16-core runner, or $336+/month for their GPU runner. And you get a far more powerful GPU.
Use Cases for GPU-Enabled CI/CD
ML Model Testing
Run inference tests against your trained models as part of CI. Catch accuracy regressions before they hit production.
CUDA Compilation
Build CUDA kernels and GPU-accelerated libraries that require a real GPU to compile and link against.
Docker Image Builds
Build and push large Docker images with pre-installed model weights. Dedicated VMs have the disk space and bandwidth for multi-GB images.
Integration Testing
Run full integration test suites with GPU-dependent services. Test your inference pipeline end-to-end with real hardware.
Step-by-step tutorial
For a detailed walkthrough including PAT setup, workflow configuration, scaling, and troubleshooting, read the full guide:
Deploy Self-Hosted GitHub Actions Runners on VectorLayStop paying per minute for CI/CD
Deploy self-hosted GitHub Actions runners in under 5 minutes. GPU access, zero queue times, and flat hourly pricing. No usage caps, no surprise bills.