Skip to content

fix: version-gate cgroup preflight check on openshell version#258

Closed
ericksoa wants to merge 1 commit intomainfrom
fix/version-gate-cgroup-check
Closed

fix: version-gate cgroup preflight check on openshell version#258
ericksoa wants to merge 1 commit intomainfrom
fix/version-gate-cgroup-check

Conversation

@ericksoa
Copy link
Contributor

Supersedes #248.

Summary

Problem

PR #248 proposed removing the cgroup check entirely since OpenShell v0.0.8 handles it upstream. But users on older openshell versions (e.g. 0.0.6) would hit cryptic k3s failures on cgroup v2 hosts (Ubuntu 24.04, DGX Spark) with no guidance on how to fix it — we saw this today on an H100.

Fix

Parse openshell --version, compare against minimum 0.0.8. If older, run the existing checkCgroupConfig() check. If current, skip it.

Keeps bin/lib/preflight.js and test/preflight.test.js intact (PR #248 deleted them).

Test plan

  • node --test test/preflight.test.js — 13/13 pass
  • node -e "require('./bin/lib/onboard')" — no import errors
  • Local openshell is 0.1.0 — cgroup check correctly skipped

Only run the cgroup v2 / cgroupns=host check for openshell < 0.0.8.
OpenShell v0.0.8 added internal cgroup v2 host namespace support
(NVIDIA/OpenShell#329), making the NemoClaw-side check unnecessary
for current versions.

For users on older openshell, the check still runs and now
recommends upgrading openshell as the primary fix.
@ericksoa
Copy link
Contributor Author

Closing in favor of #248 — using the cgroup check removal as a forcing function to get users onto openshell v0.0.8+.

@ericksoa ericksoa closed this Mar 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant