Skip to content

Conversation

@nealvaidya
Copy link
Contributor

@nealvaidya nealvaidya commented Jul 24, 2025

Add an example of vllm disaggregated serving with some explanation of what's going on

Summary by CodeRabbit

  • Documentation
    • Added a comprehensive README example explaining Dynamo's Prefill/Decode Disaggregated Serving architecture for large language model inference, including setup instructions, component roles, request flow diagrams, and operational guidance.

@nealvaidya nealvaidya marked this pull request as ready for review July 28, 2025 17:58
@nealvaidya nealvaidya requested review from a team, nnshah1 and whoisj as code owners July 28, 2025 17:58
@coderabbitai
Copy link
Contributor

coderabbitai bot commented Jul 28, 2025

Walkthrough

A new, detailed README file was added under the examples directory to demonstrate and document Dynamo's Prefill/Decode Disaggregated Serving architecture for large language model inference. The documentation explains system components, prerequisites, operational flow, deployment instructions, and architectural diagrams, but does not introduce or modify any code.

Changes

Cohort / File(s) Change Summary
Documentation Example: Disaggregated Serving
examples/basics/disaggregated_serving/README.md
Added a comprehensive README example detailing Dynamo's Prefill/Decode Disaggregated Serving architecture, including system overview, component roles, deployment steps, operational flow, and architectural diagrams. No code changes were made.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Possibly related PRs

Poem

In Dynamo’s warren, new docs appear,
Disaggregated serving, now crystal clear!
Prefill and decode, each with their role,
NIXL moves caches, making it whole.
With diagrams and steps, the path’s well-lit—
Even a bunny could deploy it! 🐇✨

Note

⚡️ Unit Test Generation is now available in beta!

Learn more here, or try it out under "Finishing Touches" below.


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
🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

  • Review comments: Directly reply to a review comment made by CodeRabbit. Example:
    • I pushed a fix in commit <commit_id>, please review it.
    • Explain this complex logic.
    • Open a follow-up GitHub issue for this discussion.
  • Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
    • @coderabbitai explain this code block.
    • @coderabbitai modularize this function.
  • PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
    • @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
    • @coderabbitai read src/utils.ts and explain its main purpose.
    • @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
    • @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

  • @coderabbitai pause to pause the reviews on a PR.
  • @coderabbitai resume to resume the paused reviews.
  • @coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
  • @coderabbitai full review to do a full review from scratch and review all the files again.
  • @coderabbitai summary to regenerate the summary of the PR.
  • @coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
  • @coderabbitai resolve resolve all the CodeRabbit review comments.
  • @coderabbitai configuration to show the current CodeRabbit configuration for the repository.
  • @coderabbitai help to get help.

Other keywords and placeholders

  • Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
  • Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
  • Add @coderabbitai anywhere in the PR title to generate the title automatically.

Documentation and Community

  • Visit our Documentation for detailed information on how to use CodeRabbit.
  • Join our Discord Community to get help, request features, and share feedback.
  • Follow us on X/Twitter for updates and announcements.

Copy link
Contributor

@coderabbitai coderabbitai bot left a comment

Choose a reason for hiding this comment

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

Actionable comments posted: 0

🧹 Nitpick comments (5)
examples/basics/disaggregated_serving/README.md (5)

40-54: Mermaid front-matter prevents GitHub rendering

GitHub strips diagrams when a mermaid code block begins with YAML front-matter (---).
Remove lines 41-43 or move the title: line inside a %%{init:...}%% comment so the diagram renders in the GitHub UI.

-```mermaid
----
-title: Disaggregated Request Flow
----
+```mermaid
+%%{init: {"theme": "default"}}%%

65-67: Explicitly mark the worker role for clarity

The decode worker is launched with the default flags. Add --role decode (or the documented equivalent) so readers don’t wonder whether the same command can accidentally start a prefill worker.

CUDA_VISIBLE_DEVICES=0 python -m dynamo.vllm \
  --role decode \
  --model Qwen/Qwen3-0.6B

80-82: Flag consistency: --is-prefill-worker vs --role prefill

If the CLI recently standardised on --role, prefer the same flag here; otherwise document both forms to avoid confusion.

CUDA_VISIBLE_DEVICES=1 python -m dynamo.vllm \
  --role prefill \
  --model Qwen/Qwen3-0.6B

170-175: Add reference link to NIXL

Readers unfamiliar with NVIDIA NIXL have no easy way to learn more.
Add an external link to the official docs or white-paper after the first mention to improve discoverability.

The architecture relies on NVIDIA's [NIXL](https://developer.nvidia.com/...link...) ...

213-216: Minor wording: avoid double “workers”

“workers distributed across multiple machines for larger scale”

Consider rephrasing slightly for readability:

“workers distributed across multiple machines to scale beyond a single node”

📜 Review details

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 13d3cc1 and 612b45a.

📒 Files selected for processing (1)
  • examples/basics/disaggregated_serving/README.md (1 hunks)
🔇 Additional comments (2)
examples/basics/disaggregated_serving/README.md (2)

30-32: Confirm the Compose file path

deploy/metrics/docker-compose.yml looks like an infra-metrics stack rather than the etcd + NATS stack referenced in the text.
Double-check that the Compose file actually starts both etcd and NATS (and nothing extra) or point to the correct compose file (e.g. deploy/core/docker-compose.yml).
Otherwise readers may start the wrong services.


36-39: The existing relative links are correct – no change needed
The links use ../../../components/... from examples/basics/disaggregated_serving/README.md, which resolves to components/... at the repo root. The reviewer’s suggested ../../../../ overshoots and would break the paths.
Please disregard the suggested update.

Likely an incorrect or invalid review comment.

@nealvaidya nealvaidya requested a review from ishandhanani July 28, 2025 23:47
Copy link
Collaborator

@whoisj whoisj left a comment

Choose a reason for hiding this comment

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

Approved w/ a comment\question.

@nealvaidya nealvaidya enabled auto-merge (squash) July 29, 2025 17:43
@nealvaidya nealvaidya merged commit 291df28 into main Jul 29, 2025
8 of 9 checks passed
@nealvaidya nealvaidya deleted the nealv/disagg_example branch July 29, 2025 18:35
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants