[Test] Drop remote --headless for vLLM 0.28 hybrid DP nightlies by LQDLove · Pull Request #16133 · vllm-project/vllm-ascend

LQDLove · 2026-09-09T07:01:54Z

What this PR does / why we need it?

vLLM 0.28 infers hybrid LB whenever --data-parallel-start-rank is set (data_parallel_start_rank is not None). In that mode the DP handshake requires remote engines not to pass --headless:

Remote engine must not use --headless in external or hybrid dp lb mode

Nightly double-node / multi-node jobs still used the old internal-LB remote command (--data-parallel-start-rank + --headless), so startup was rejected during handshake.

Drop --headless on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). Affected nightly YAMLs:

tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_198k_function.yaml
tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_128k_90_50.yaml
tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml
tests/e2e/nightly/multi_node/internal_dp/config/GLM5_2-W8A8-A3-dual-nodes.yaml
tests/e2e/nightly/multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml

Does this PR introduce any user-facing change?

No.

How was this patch tested?

YAML-only change. Inspected vLLM main EngineArgs.create_engine_config and vllm/v1/engine/utils.py handshake (remote_should_be_headless).
Nightly CI should re-run the five jobs above.
vLLM main: vllm-project/vllm@b2f6858

vLLM 0.28 infers hybrid LB from --data-parallel-start-rank, and that mode rejects --headless on remote nodes during handshake. Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com>

gemini-code-assist · 2026-09-09T07:02:07Z

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request updates the nightly test configurations for GLM-5.1 models to resolve compatibility issues with vLLM 0.28. By removing the --headless flag in hybrid data-parallel mode, the remote engines can now correctly participate in the handshake process, ensuring stable execution in multi-node environments.

Highlights

Configuration Update: Removed the --headless flag from GLM-5.1 hybrid data-parallel configurations to comply with vLLM 0.28 requirements.

New Features

🧠 You can now enable Memory (public preview) to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize the Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counterproductive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

gemini-code-assist

Code Review

This pull request removes the --headless flag from the deployment configurations in two GLM-5.1 multi-node test YAML files. The review feedback correctly identifies that the PR Title and Summary do not comply with the repository's style guide and provides suggestions to align them with the required format.

gemini-code-assist · 2026-09-09T07:02:59Z

      --data-parallel-size 8
      --data-parallel-size-local 4
      --data-parallel-start-rank 4
-      --headless


The Pull Request Title and Summary do not fully adhere to the repository style guide. Specifically, the PR Title is missing the [Action] prefix (e.g., [BugFix]), and the PR Summary headers should use H3 (###) instead of H2 (##).

Please update the PR Title and Summary to match the following suggestions:

Suggested PR Title:

[Test][BugFix] Drop remote --headless for GLM-5.1 hybrid DP

Suggested PR Summary:

### What this PR does / why we need it? vLLM 0.28 infers hybrid LB whenever `--data-parallel-start-rank` is set. In that mode, the DP handshake rejects remote engines that still pass `--headless` with the error: `Remote engine must not use --headless in external or hybrid dp lb mode` This failed the nightly `double-node` test for `GLM-5.1-W8A8C8-A3_198k_function`. This PR drops `--headless` on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). The same leftover flag is also removed from the sibling 128k nightly YAML. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? - Inspected vLLM main `EngineArgs.create_engine_config` (`data_parallel_start_rank is not None` implies hybrid LB) and `vllm/v1/engine/utils.py` handshake (`remote_should_be_headless`). - YAML-only change; nightly double-node CI should re-run `GLM-5.1-W8A8C8-A3_198k_function`.

References

The PR Title and PR Summary must follow the specified format, with the PR Title using [Branch][Module][Action] format and the PR Summary using specific H3 headers inside markdown code blocks. ^(link)

github-actions · 2026-09-09T07:03:32Z

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

A PR should do only one thing, smaller PRs enable faster reviews.
Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

Same vLLM 0.28 handshake reject as GLM-5.1 C8: start-rank implies hybrid LB, so GLM-5.1/5.2 A3 dual-node and Kimi-K2.5 A2 remote nodes cannot pass --headless. Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com>

czydyy · 2026-09-09T07:25:28Z

/nightly GLM5_1-W8A8-A3-dual-nodes Kimi-K2_5-W4A8-A2-dual-nodes
nightly command triggered.

Ronald1995

LGTM

…-project#16133) ### What this PR does / why we need it? vLLM 0.28 infers hybrid LB whenever `--data-parallel-start-rank` is set (`data_parallel_start_rank is not None`). In that mode the DP handshake requires remote engines **not** to pass `--headless`: ```text Remote engine must not use --headless in external or hybrid dp lb mode ``` Nightly double-node / multi-node jobs still used the old internal-LB remote command (`--data-parallel-start-rank` + `--headless`), so startup was rejected during handshake. Drop `--headless` on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). Affected nightly YAMLs: - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_198k_function.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_128k_90_50.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_2-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? - YAML-only change. Inspected vLLM main `EngineArgs.create_engine_config` and `vllm/v1/engine/utils.py` handshake (`remote_should_be_headless`). - Nightly CI should re-run the five jobs above. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com>

### What this PR does / why we need it? vLLM #47692 changed the hybrid DP load-balancing inference to treat an explicit `--data-parallel-start-rank 0` as meaningful. As a result, the primary node in these multi-node `internal_dp` configurations is now inferred as hybrid LB, while the remote node is intended to run headless for internal LB. PRs #16133, #16186, and #16204 removed `--headless` from remote nodes to satisfy the resulting handshake, but that changed the topology covered by the tests. The regular multi-node internal-DP runner sends benchmark traffic only to the primary node. In disaggregated-prefill cases, the PD proxy was explicitly designed to exclude headless nodes and target one API endpoint per DP group. Switching these deployments to hybrid/external LB therefore changes the intended test coverage. Restore the intended internal-LB topology in all 17 affected nightly and weekly configurations: - remove `--data-parallel-start-rank 0` from the primary node of each DP group, allowing rank 0 to be inferred without enabling hybrid LB; - restore `--headless` on the remote node while retaining its non-zero DP rank offset. This keeps one API endpoint on the primary node and lets it schedule requests across all local and remote DP ranks. ### Does this PR introduce _any_ user-facing change? No. This only updates nightly test deployment configurations. ### How was this patch tested? - `git diff --check` - Parsed all 17 modified YAML files with PyYAML and verified their internal-DP node roles. - Multi-node NPU nightly and weekly jobs should validate the complete deployments. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Co-authored-by: jiangkaiqiang <jiangkaiqiang@huawei.com>

…-project#16133) ### What this PR does / why we need it? vLLM 0.28 infers hybrid LB whenever `--data-parallel-start-rank` is set (`data_parallel_start_rank is not None`). In that mode the DP handshake requires remote engines **not** to pass `--headless`: ```text Remote engine must not use --headless in external or hybrid dp lb mode ``` Nightly double-node / multi-node jobs still used the old internal-LB remote command (`--data-parallel-start-rank` + `--headless`), so startup was rejected during handshake. Drop `--headless` on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). Affected nightly YAMLs: - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_198k_function.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_128k_90_50.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_2-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? - YAML-only change. Inspected vLLM main `EngineArgs.create_engine_config` and `vllm/v1/engine/utils.py` handshake (`remote_should_be_headless`). - Nightly CI should re-run the five jobs above. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com>

### What this PR does / why we need it? vLLM #47692 changed the hybrid DP load-balancing inference to treat an explicit `--data-parallel-start-rank 0` as meaningful. As a result, the primary node in these multi-node `internal_dp` configurations is now inferred as hybrid LB, while the remote node is intended to run headless for internal LB. PRs vllm-project#16133, vllm-project#16186, and vllm-project#16204 removed `--headless` from remote nodes to satisfy the resulting handshake, but that changed the topology covered by the tests. The regular multi-node internal-DP runner sends benchmark traffic only to the primary node. In disaggregated-prefill cases, the PD proxy was explicitly designed to exclude headless nodes and target one API endpoint per DP group. Switching these deployments to hybrid/external LB therefore changes the intended test coverage. Restore the intended internal-LB topology in all 17 affected nightly and weekly configurations: - remove `--data-parallel-start-rank 0` from the primary node of each DP group, allowing rank 0 to be inferred without enabling hybrid LB; - restore `--headless` on the remote node while retaining its non-zero DP rank offset. This keeps one API endpoint on the primary node and lets it schedule requests across all local and remote DP ranks. ### Does this PR introduce _any_ user-facing change? No. This only updates nightly test deployment configurations. ### How was this patch tested? - `git diff --check` - Parsed all 17 modified YAML files with PyYAML and verified their internal-DP node roles. - Multi-node NPU nightly and weekly jobs should validate the complete deployments. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Co-authored-by: jiangkaiqiang <jiangkaiqiang@huawei.com>

…-project#16133) ### What this PR does / why we need it? vLLM 0.28 infers hybrid LB whenever `--data-parallel-start-rank` is set (`data_parallel_start_rank is not None`). In that mode the DP handshake requires remote engines **not** to pass `--headless`: ```text Remote engine must not use --headless in external or hybrid dp lb mode ``` Nightly double-node / multi-node jobs still used the old internal-LB remote command (`--data-parallel-start-rank` + `--headless`), so startup was rejected during handshake. Drop `--headless` on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). Affected nightly YAMLs: - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_198k_function.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_128k_90_50.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_2-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? - YAML-only change. Inspected vLLM main `EngineArgs.create_engine_config` and `vllm/v1/engine/utils.py` handshake (`remote_should_be_headless`). - Nightly CI should re-run the five jobs above. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com> Signed-off-by: tianming2009 <13246728590@163.com>

### What this PR does / why we need it? vLLM #47692 changed the hybrid DP load-balancing inference to treat an explicit `--data-parallel-start-rank 0` as meaningful. As a result, the primary node in these multi-node `internal_dp` configurations is now inferred as hybrid LB, while the remote node is intended to run headless for internal LB. PRs vllm-project#16133, vllm-project#16186, and vllm-project#16204 removed `--headless` from remote nodes to satisfy the resulting handshake, but that changed the topology covered by the tests. The regular multi-node internal-DP runner sends benchmark traffic only to the primary node. In disaggregated-prefill cases, the PD proxy was explicitly designed to exclude headless nodes and target one API endpoint per DP group. Switching these deployments to hybrid/external LB therefore changes the intended test coverage. Restore the intended internal-LB topology in all 17 affected nightly and weekly configurations: - remove `--data-parallel-start-rank 0` from the primary node of each DP group, allowing rank 0 to be inferred without enabling hybrid LB; - restore `--headless` on the remote node while retaining its non-zero DP rank offset. This keeps one API endpoint on the primary node and lets it schedule requests across all local and remote DP ranks. ### Does this PR introduce _any_ user-facing change? No. This only updates nightly test deployment configurations. ### How was this patch tested? - `git diff --check` - Parsed all 17 modified YAML files with PyYAML and verified their internal-DP node roles. - Multi-node NPU nightly and weekly jobs should validate the complete deployments. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Co-authored-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Signed-off-by: tianming2009 <13246728590@163.com>

…-project#16133) ### What this PR does / why we need it? vLLM 0.28 infers hybrid LB whenever `--data-parallel-start-rank` is set (`data_parallel_start_rank is not None`). In that mode the DP handshake requires remote engines **not** to pass `--headless`: ```text Remote engine must not use --headless in external or hybrid dp lb mode ``` Nightly double-node / multi-node jobs still used the old internal-LB remote command (`--data-parallel-start-rank` + `--headless`), so startup was rejected during handshake. Drop `--headless` on the remote node so both nodes run hybrid LB (each exposes an API and owns its local ranks). Affected nightly YAMLs: - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_198k_function.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM-5.1-W8A8C8-A3_128k_90_50.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/GLM5_2-W8A8-A3-dual-nodes.yaml` - `tests/e2e/nightly/multi_node/internal_dp/config/Kimi-K2_5-W4A8-A2-dual-nodes.yaml` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? - YAML-only change. Inspected vLLM main `EngineArgs.create_engine_config` and `vllm/v1/engine/utils.py` handshake (`remote_should_be_headless`). - Nightly CI should re-run the five jobs above. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com> Signed-off-by: like-0517 <ithwlike@126.com>

### What this PR does / why we need it? vLLM #47692 changed the hybrid DP load-balancing inference to treat an explicit `--data-parallel-start-rank 0` as meaningful. As a result, the primary node in these multi-node `internal_dp` configurations is now inferred as hybrid LB, while the remote node is intended to run headless for internal LB. PRs vllm-project#16133, vllm-project#16186, and vllm-project#16204 removed `--headless` from remote nodes to satisfy the resulting handshake, but that changed the topology covered by the tests. The regular multi-node internal-DP runner sends benchmark traffic only to the primary node. In disaggregated-prefill cases, the PD proxy was explicitly designed to exclude headless nodes and target one API endpoint per DP group. Switching these deployments to hybrid/external LB therefore changes the intended test coverage. Restore the intended internal-LB topology in all 17 affected nightly and weekly configurations: - remove `--data-parallel-start-rank 0` from the primary node of each DP group, allowing rank 0 to be inferred without enabling hybrid LB; - restore `--headless` on the remote node while retaining its non-zero DP rank offset. This keeps one API endpoint on the primary node and lets it schedule requests across all local and remote DP ranks. ### Does this PR introduce _any_ user-facing change? No. This only updates nightly test deployment configurations. ### How was this patch tested? - `git diff --check` - Parsed all 17 modified YAML files with PyYAML and verified their internal-DP node roles. - Multi-node NPU nightly and weekly jobs should validate the complete deployments. - vLLM main: vllm-project/vllm@b2f6858 --------- Signed-off-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Co-authored-by: jiangkaiqiang <jiangkaiqiang@huawei.com> Signed-off-by: like-0517 <ithwlike@126.com>

[Test] Drop remote --headless for GLM-5.1 hybrid DP

8604eba

vLLM 0.28 infers hybrid LB from --data-parallel-start-rank, and that mode rejects --headless on remote nodes during handshake. Signed-off-by: LQD <1107297340@qq.com> Co-authored-by: Cursor <cursoragent@cursor.com>

gemini-code-assist Bot reviewed Sep 9, 2026

View reviewed changes

github-actions Bot added the module:tests label Sep 9, 2026

LQDLove changed the title ~~[Test] Drop remote --headless for GLM-5.1 hybrid DP~~ [Test] Drop remote --headless for vLLM 0.28 hybrid DP nightlies Sep 9, 2026

zhao-stack added the ready-precise run selected e2e test for pr label Sep 9, 2026

Ronald1995 approved these changes Sep 9, 2026

View reviewed changes

wenjun91 approved these changes Sep 9, 2026

View reviewed changes

wenjun91 merged commit 7cb184e into vllm-project:main Sep 9, 2026
28 checks passed

LQDLove deleted the fix/glm51-hybrid-lb-drop-headless branch September 10, 2026 00:53

DavidJiang9 mentioned this pull request Sep 11, 2026

[Test] Restore internal DP for multi-node tests #16330

Merged

Navigation Menu

Repository navigation

[Test] Drop remote --headless for vLLM 0.28 hybrid DP nightlies - #16133

[Test] Drop remote --headless for vLLM 0.28 hybrid DP nightlies#16133
wenjun91 merged 2 commits into
vllm-project:mainfrom
LQDLove:fix/glm51-hybrid-lb-drop-headless

LQDLove commented Sep 9, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot commented Sep 9, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

gemini-code-assist Bot Sep 9, 2026

Uh oh!

github-actions Bot commented Sep 9, 2026

Uh oh!

czydyy commented Sep 9, 2026 •

edited by vllm-ascend-ci

Loading

Uh oh!

Ronald1995 left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

LQDLove commented Sep 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this PR does / why we need it?

Does this PR introduce any user-facing change?

How was this patch tested?

Uh oh!

gemini-code-assist Bot commented Sep 9, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Sep 9, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Sep 9, 2026

Uh oh!

czydyy commented Sep 9, 2026 • edited by vllm-ascend-ci Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Ronald1995 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

LQDLove commented Sep 9, 2026 •

edited

Loading

czydyy commented Sep 9, 2026 •

edited by vllm-ascend-ci

Loading