GPU-Extender

GPU-Extender is a Kubernetes scheduler extender that allocates and shares GPU capacity across cluster nodes, improving utilization for graphics-intensive…

  • Infrastructure
  • Tools & Utilities
GPU-Extender is a Kubernetes scheduler extender that allocates and shares GPU capacity across cluster nodes, improving utilization for graphics-intensive workloads.

Platform availability

Available on 0 of 10 platforms

  • Unraid (not listed)
  • TrueNAS (not listed)
  • Umbrel (not listed)
  • ZimaOS (not listed)
  • Proxmox (not listed)
  • Helm (not listed)
  • Mac (not listed)
  • Windows (not listed)
  • Linux (not listed)
  • Any server (not listed)

Health score?A 0-100 score computed nightly from four factors: maintained (40%: recent commits and steady activity), popular (25%: GitHub stars, log scale), easy to install (20%: platforms, install notes, setup guides, Docker image) and light to run (15%: minimum RAM, ARM support). When the source code has been checked, a fifth factor, security check (20%: automated scan for committed secrets, unsafe desktop-app settings, known-vulnerable dependencies and project hygiene; any serious finding caps it at 20), is added. Factors without data are left out and the rest are rescaled.

30/100 Low

Maintained
0last commit 1346 days ago
Popular
641,530 GitHub stars
Easy to install
250 install platforms, install notes, Docker image
Light to run
No data
Security check
54no serious findings, 4 minor, CI, OpenSSF Scorecard 3.5/10

Based on 4 of 5 factors.

GitHub stars
1,530
Open issues
111
Last commit
2023-03-16
License
Apache-2.0
Activity
stale

Checked yesterday - source: GitHub (AliyunContainerService/gpushare-scheduler-extender)

Resources & Compatibility

ARM (e.g. Raspberry Pi) is not supported.

Checked 54 days ago - source: Docker Hub image tags and published docs

First-install notes

  • Install Helm before deploying the GPU operator Helm (a tool that deploys Kubernetes apps from packaged charts) needs to be on your machine before you can install or configure the GPU operator. If it’s missing, the Helm commands below will fail and you won’t be able to start the install. source
  • Get GKE credentials if using Google Kubernetes If your cluster is Google Kubernetes Engine (GKE), run the gcloud (Google’s command-line tool) get-credentials command for your cluster first. Without local cluster credentials, kubectl and Helm can’t log in, so the install can’t be applied. source
  • Check that the GPU runtime classes appear After install, run kubectl (the command-line tool for talking to a Kubernetes cluster) get runtimeclass and look for NVIDIA/Kata entries such as kata-qemu-nvidia-gpu. This confirms the GPU runtime options are registered; if they’re missing, workloads that request a GPU runtime may be stuck or rejected. source

Fetched 53 days ago - each note links to its source

Alternatives

No alternatives collected yet.

Common questions

What Docker configuration is required before deploying GPU-Extender?

Docker must be configured with Nvidia as the default runtime. The NVIDIA driver and nvidia-docker prerequisites must also be in place.

Do I need to install a separate GPU device plugin with the scheduler extender?

Yes, the GPU sharing solution uses both scheduler extenders and the device plugin mechanism. The device plugin is maintained as a separate project.

How can I inspect GPU sharing state from kubectl?

Build the kubectl-inspect-gpushare-v2 binary from the gpushare-device-plugin source. It is intended as a kubectl extension for inspecting GPU sharing.

Can GPU-Extender be installed with Helm?

For Alibaba Cloud Kubernetes, the docs point to a deployer README for installing with Helm Charts. Otherwise the README points to the project installation guide.

Can multiple GPU-sharing pods be scheduled onto the same GPU device?

The project demos deploying multiple GPU shared pods and scheduling them on the same GPU device in a binpack way.

Answers sourced from github.com

Community

Collected 48 days ago - public community threads