Skip to content

doc: add doc best-practice/deploy-inference-service#386

Open
diw-zw wants to merge 3 commits into
sgl-project:mainfrom
diw-zw:0710-doc-01
Open

doc: add doc best-practice/deploy-inference-service#386
diw-zw wants to merge 3 commits into
sgl-project:mainfrom
diw-zw:0710-doc-01

Conversation

@diw-zw

@diw-zw diw-zw commented Jul 10, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

@gemini-code-assist

Copy link
Copy Markdown

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@cheyang

cheyang commented Jul 13, 2026

Copy link
Copy Markdown
Collaborator

/copilot review

@cheyang

cheyang commented Jul 13, 2026

Copy link
Copy Markdown
Collaborator

@gemini-code-assist /review

@gemini-code-assist

Copy link
Copy Markdown

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@cheyang

cheyang commented Jul 13, 2026

Copy link
Copy Markdown
Collaborator

Thanks for putting together this best-practice guide -- the multi-role and topology concepts are clearly explained with practical examples, and the operations guide is well-structured with verification steps.

A few things to address before merge:

  1. Mixed-language text in doc/best-practice/en/01-deploy-inference-service.md line 24: the sentence contains Chinese characters (承担) in English text -- looks like a translation leftover.
  2. Incorrect label selector in the "Verify Deployment" section (both EN and ZH): kubectl get pods -l app=llm-inference uses a label that doesn't appear in any YAML example. Should be rbg.workloads.x-k8s.io/group-name=<rbg-name>.
  3. Broken links in Related Documents: 02-using-role-templates.md, 08-configuring-autoscaling.md, 03-configuring-rolling-updates.md don't exist yet.

Items 1 and 2 are straightforward fixes. Once addressed, this should be good to go.

---

## Multi-Role Definition
The core concept of RBG is the **Role**. An inference service can consist of one or more roles, each承担ing different responsibilities. Depending on the inference architecture, the number and division of roles varies.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The sentence "each承担ing different responsibilities" contains Chinese characters in what should be pure English text. This looks like a leftover from translation. Suggest: "each taking on different responsibilities".

kubectl get rbg <rbg-name> -o wide

# Check Pod status
kubectl get pods -l app=llm-inference

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This command uses app=llm-inference which is never defined in any YAML example in this document. The correct label (as used in the operations guide) is rbg.workloads.x-k8s.io/group-name=<rbg-name>. Users following this will get empty results.

Suggested fix:

kubectl get pods -l rbg.workloads.x-k8s.io/group-name=<rbg-name>

kubectl get rbg <rbg-name> -o wide

# 查看 Pod 状态
kubectl get pods -l app=llm-inference

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same issue as the English version: app=llm-inference is never defined in any YAML example. Should use rbg.workloads.x-k8s.io/group-name=<rbg-name> to be consistent with the operations guide.

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.

2 participants