Skip to content

docs: Add IPv6 network pages for standalone DCGM - #516

Merged
mikemckiernan merged 2 commits into
mainfrom
devin/1790632390-standalone-dcgm-ipv6
Oct 7, 2026
Merged

mikemckiernan merged 2 commits into
mainfrom
devin/1790632390-standalone-dcgm-ipv6

Conversation

@mikemckiernan

@mikemckiernan mikemckiernan commented Sep 28, 2026 •

Copy link
Copy Markdown
Member

Summary

This PR adds IPv6 network pages to both docsets. DCGM is the only component that needs IPv6-specific configuration, and only when it runs as a standalone daemon set: DCGM Exporter then connects to the nvidia-dcgm service on port 5555.

  • gpu-operator/install-gpu-operator-ipv6.rst, under Specialized Networks. Helm defaults to dcgm.enabled: false, so most readers need no IPv6 configuration. The page covers the opt-in standalone case.
  • openshift/deploy-gpu-ocp-ipv6.rst. On OpenShift, the default ClusterPolicy from the CSV alm-examples sets dcgm.enabled: true. Source: the published v26.7.1 bundle, operators/gpu-operator-certified/26.7.1/manifests/gpu-operator-certified.clusterserviceversion.yaml in redhat-openshift-ecosystem/certified-operators (commit 8fd2c5e). So every IPv6-only OCP cluster needs the dcgm.args override.

The standalone case needs two things:

  • An IPv6 bind for the host engine. The DCGM image's default CMD is -n -b 0.0.0.0 --log-level NONE -f -, which listens on IPv4 only. The Operator sets no args by default, and a non-empty dcgm.args replaces the container args entirely (controllers/object_controls.go). So the pages show the full list:
    dcgm:
      enabled: true
      args: ["-n", "-b", "::", "--log-level", "NONE", "-f", "-"]
  • GPU Operator v26.7.1 or later. DCGM 4.6.1 changed DcgmIpc::ConnectTcpAsyncImpl to resolve hostnames with AF_UNSPEC instead of forcing AF_INET. Before that, the exporter couldn't resolve nvidia-dcgm to an IPv6 address. Exporter 4.6.1-4.8.4, which ships in v26.7.1, bundles that client library.

I deliberately left out a cross-link from the troubleshooting pages to keep this content out of the way for most readers.

The docs build wasn't run locally, at the reviewer's request. Pre-commit hooks passed. The only other change in gpu-operator/index.rst is a trailing-whitespace fix made by the pre-commit hook.

Link to Devin session: https://nvidia-cloud.devinenterprise.com/sessions/ec15cc48cd754a1eae4ef22eec691b5d
Open in Devin Desktop: https://nvidia-cloud.devinenterprise.com/desktop/session/ec15cc48cd754a1eae4ef22eec691b5d?variant=devin
Requested by: @mikemckiernan

Signed-off-by: Mike McKiernan <mmckiernan@nvidia.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
@devin-ai-integration

Copy link
Copy Markdown
Contributor

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

  • Disable automatic comment, CI, and merge conflict monitoring

@github-actions

Copy link
Copy Markdown

Documentation preview

https://nvidia.github.io/cloud-native-docs/review/pr-516

Signed-off-by: Mike McKiernan <mmckiernan@nvidia.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
@devin-ai-integration devin-ai-integration Bot changed the title docs: Add IPv6 networks page for standalone DCGM docs: Add IPv6 network pages for standalone DCGM Sep 28, 2026
@mikemckiernan mikemckiernan added the documentation Issue/PR focused on fixing/editing/adding documentation bits label Oct 7, 2026
@francisguillier

Copy link
Copy Markdown
Contributor

thanks @mikemckiernan
the changes LGTM.

@devin-ai-integration

Copy link
Copy Markdown
Contributor

Devin is archived and cannot be woken up. Please unarchive Devin if you want to continue using it.

View session

@mikemckiernan
mikemckiernan merged commit e2861bc into main Oct 7, 2026
2 checks passed
@mikemckiernan
mikemckiernan deleted the devin/1790632390-standalone-dcgm-ipv6 branch October 7, 2026 19:40
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Issue/PR focused on fixing/editing/adding documentation bits

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants