NVIDIA / NVIDIA/OpenShell

Improve OpenShell Support for HPC-style Shared Compute Environments

Open
#1,749 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

state:stale
Dominant language
Rust
Stars
8.7k
Forks
1.3k
Avg merge
2d 11h
Merged PRs (30d)
253

Description

Problem Statement
Summary

This umbrella issue tracks a set of related feature requests to make OpenShell easier to deploy in HPC-style shared compute environments.

The common theme is support for locked-down Linux hosts, shared filesystems, scheduler-managed jobs, narrow internal network access, and daemonless runtime options.

This umbrella supersedes #1393. #1393 raised the general problem of HPC-style hosts being a poor fit for Docker/Podman/MicroVM/Kubernetes. The work is concrete and each runtime can be evaluated, accepted, or dropped on its own merits.

Motivation

Many HPC environments differ from cloud-native environments:

  • shared filesystems are central to the workflow,
  • tools and datasets are often installed outside the sandbox image,
  • users may not have admin privileges,
  • compute nodes may not run Docker or Kubernetes,
  • network egress is often restricted,
  • batch schedulers manage job lifecycle,
  • workloads may need fast startup and low runtime overhead.

OpenShell has a strong policy, gateway, credential, and supervision model, but these environments need additional runtime and storage support.

Proposed Design
Child issues
  1. #1746 — First-class shared filesystem mounts for HPC sandboxes
  2. #1747 — Raw TCP allowlists for internal HPC services
  3. #1748 — Filesystem policy enforcement on shared-mounted paths (incl. user-identity / UID-GID model)
  4. #1393 — Bubblewrap/Apptainer-backed runtime mode for daemonless HPC hosts
Desired outcome

OpenShell should be able to run automated or agentic workloads on traditional HPC/shared compute infrastructure while preserving least-privilege policy enforcement, credential mediation, network controls, and auditability.

Alternatives Considered

N/A

Agent Investigation

No response

Checklist
  • I've reviewed existing issues and the architecture docs
  • This is a design proposal, not a "please build this" request

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reading child issues #1746, #1747, and #1748, then compare their scopes with the superseded #1393. The issue is an umbrella design proposal rather than a standalone implementation task; done means evaluating the child runtime, filesystem, and networking requests while preserving policy enforcement, credential mediation, auditability, and least privilege.

Written by the indexing model from the issue text.

Assessment

Tech stack
docker, hpc, kubernetes, linux
Domain
hpc, infrastructure, operating-systems
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.