docs: clarify GPU Memory Service status

[athreesh]

Summary

• Add a GPU Memory Service Kubernetes docs page that explains when to use GMS today, when not to use it, and how it relates to Snapshot and failover.
• Update Snapshot docs to stop telling users to enable spec.gpuMemoryService.enabled, since Snapshot plus GMS is currently rejected by admission because of GPU driver restore issues.
• Link the new page from the Kubernetes docs nav and README, and update the DynamoCheckpoint sample comment to keep GMS checkpoint capture disabled.

Why

The current Snapshot docs made GMS checkpoint/restore look user-ready even though the operator now blocks that path. The new page makes the experimental status explicit and gives users a clearer decision point for GMS, Snapshot, and failover.

Validation

• git diff --check
• git diff --cached --check
• ruby -e "require 'yaml'; YAML.load_file('docs/index.yml'); puts 'docs/index.yml ok'"
• verified internal doc targets exist for gpu-memory-service.md, snapshot.md, and api-reference.md

Note: fern is not installed in this local environment, so I did not run fern check or fern docs broken-links.

Summary by CodeRabbit

• Documentation

  • Added comprehensive GPU Memory Service documentation explaining same-node recovery behavior and current limitations
  • Updated navigation to include GPU Memory Service deployment guide
  • Clarified that GPU Memory Service is temporarily disabled with Snapshot due to GPU driver restore issues

• Chores

  • Updated sample configuration with guidance notes for GPU Memory Service

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Walkthrough

This PR documents the experimental GPU Memory Service (GMS) feature by adding a comprehensive new documentation page, updating related documentation to clarify GMS limitations and its temporary incompatibility with Snapshot due to GPU driver restore issues, adding navigation links, and updating a sample configuration comment to reflect that GMS checkpoint/restore is temporarily disabled.

Changes

GPU Memory Service Documentation & Integration

Layer / File(s)	Summary
Core Feature Documentation docs/kubernetes/gpu-memory-service.md	New page explaining GMS purpose (keeping GPU-resident weights across lifecycle changes on same node), failure recovery flow with diagram, decision guide, prerequisites, limitations, API placement for both v1alpha1 and v1beta1, and example manifests for basic GMS and active/passive failover usage.
Integration with Existing Docs docs/kubernetes/snapshot.md, docs/fault-tolerance/README.md	snapshot.md updated to prohibit GMS enabled with Snapshot, mark failover restore as experimental GMS-only, and note GMS restore is disabled due to driver issues. fault-tolerance/README.md gains GPU Memory entry in fault-tolerance table and dedicated GMS section clarifying same-node recovery scope and non-coverage of hardware loss, in-flight requests, and KV cache.
Navigation & Discovery docs/index.yml, docs/kubernetes/README.md	Navigation entries added for "GPU Memory Service" page in Kubernetes Deployment Guide and Additional Resources lists.
Configuration Samples deploy/operator/config/samples/nvidia.com_v1alpha1_dynamocheckpoint.yaml	Sample DynamoCheckpoint spec comment clarifies that GMS checkpoint/restore is temporarily disabled due to GPU driver restore issues and should remain false.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'docs: clarify GPU Memory Service status' directly and concisely summarizes the main objective: updating documentation to clarify the GPU Memory Service feature status and guidance.
Description check	✅ Passed	The PR description includes all required sections: summary of changes, rationale (Why), and validation performed, with clear details on what was changed and why.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

Tip

💬 Introducing Slack Agent: The best way for teams to turn conversations into code.

Slack Agent is built on CodeRabbit's deep understanding of your code, so your team can collaborate across the entire SDLC without losing context.

• Generate code and open pull requests
• Plan features and break down work
• Investigate incidents and troubleshoot customer tickets together
• Automate recurring tasks and respond to alerts with triggers
• Summarize progress and report instantly

Built for teams:

• Shared memory across your entire org—no repeating context
• Per-thread sandboxes to safely plan and execute work
• Governance built-in—scoped access, auditability, and budget controls

One agent for your entire SDLC. Right inside Slack.

👉 Get started

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@docs/kubernetes/gpu-memory-service.md`:
- Around line 21-42: The fenced ASCII diagram block in the GPU memory service
doc is missing a language identifier and triggers markdownlint MD040; update the
opening code fence for the diagram (the triple-backtick that precedes the ASCII
art block) to include a language tag such as "text" (e.g., ```text) so the
fenced code block is explicitly marked and the linter warning is resolved.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 6babb8cf-35c4-4de5-8738-090dbc250860

📥 Commits

Reviewing files that changed from the base of the PR and between a039628 and 9f52639.

📒 Files selected for processing (6)

• deploy/operator/config/samples/nvidia.com_v1alpha1_dynamocheckpoint.yaml
• docs/fault-tolerance/README.md
• docs/index.yml
• docs/kubernetes/README.md
• docs/kubernetes/gpu-memory-service.md
• docs/kubernetes/snapshot.md

[athreesh]

/ok to test 5aeec39

[github-actions]

🌿 Fern Docs Preview: https://nvidia-preview-b958f686-fc62-4dbd-b276-c69d88df5d9f.docs.buildwithfern.com/dynamo/dev

[athreesh]

/ok to test 74825fb