ModelServing Role Support LeaderWorkerSet #609

zhiweideren · 2025-12-30T09:43:17Z

What type of PR is this?
enhancement

What this PR does / why we need it:
This proposal introduces LeaderWorkerSet (LWS) role support for Kthena ModelServing. The goal is to enable ModelServing to orchestrate leader/worker style “super-pod” workloads via LWS where it provides better lifecycle semantics (group-level rollout, coordinated restart behaviors, stable identity and networking patterns), while preserving the existing ModelServing user experience and integrations (ModelServer/ModelRoute selection and autoscaling).

Which issue(s) this PR fixes:
Fixes #407

Special notes for your reviewer:
Determine which solution to use

Does this PR introduce a user-facing change?:

gemini-code-assist · 2025-12-30T09:43:32Z

Summary of Changes

Hello @zhiweideren, 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 proposes an enhancement to Kthena ModelServing by integrating LeaderWorkerSet (LWS) support. The primary goal is to enable ModelServing to effectively orchestrate leader/worker style "super-pod" workloads, benefiting from LWS's superior lifecycle management, coordinated restarts, and stable identity. The proposal evaluates various approaches and recommends "Extended Role Configuration," which allows specific ModelServing roles to be optionally backed by LWS resources, ensuring existing ModelServing routing and autoscaling integrations remain functional.

Highlights

LeaderWorkerSet Support: Introduces support for LeaderWorkerSet (LWS) within Kthena ModelServing to orchestrate leader/worker "super-pod" workloads.
Enhanced Lifecycle Semantics: Leverages LWS for improved group-level rollout, coordinated restart behaviors, and stable identity/networking patterns for these workloads.
Preserved User Experience: Ensures existing ModelServing user experience and integrations (ModelServer/ModelRoute selection and autoscaling) are maintained.
Recommended Solution: Adopts "Solution 3: Extended Role Configuration," which extends ModelServing.spec.template.roles[] to optionally be backed by LWS resources, allowing mixed deployments.
Clear Migration Path: Provides a clear path for existing ModelServing roles to opt into LWS without altering higher-level ModelServing constructs.

🧠 New Feature in Public Preview: You can now enable Memory 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 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 counter productive. 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.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

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 introduces a design proposal for adding LeaderWorkerSet (LWS) support to Kthena ModelServing. The proposal is well-structured, comprehensive, and clearly evaluates multiple solutions before recommending one. The recommended approach (Solution 3: Extended Role Configuration) seems sound and balances user experience with implementation complexity.

My feedback focuses on adding more specific details to the proposed API changes and clarifying the behavior for certain lifecycle operations, such as migrating a role between being LWS-backed and natively managed. These clarifications will help solidify the design and prevent potential user confusion.

docs/proposal/modelserving-role-support-leaderworkerset.md

Copilot

Pull request overview

This PR introduces LeaderWorkerSet (LWS) role support for Kthena ModelServing, enabling users to submit LeaderWorkerSet Custom Resources that are automatically converted to ModelServing internal resources. The implementation creates a new LWS controller that watches LeaderWorkerSet CRs and maps them to ModelServing resources without requiring the native LWS controller to be deployed.

Key Changes

Adds a new LWS controller that translates LeaderWorkerSet specifications to ModelServing resources
Integrates LWS controller startup into the main controller lifecycle
Includes comprehensive proposal documentation explaining the architecture and mapping rules

Reviewed changes

Copilot reviewed 9 out of 10 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
pkg/model-serving-controller/controller/lws_controller.go	New controller implementing LWS to ModelServing conversion and reconciliation logic
pkg/controller/controller.go	Integration point for starting the LWS controller when ModelServing controller is enabled
docs/proposal/modelserving-role-support-leaderworkerset.md	Detailed proposal document describing architecture, field mappings, and implementation requirements
go.mod	Adds sigs.k8s.io/lws v0.7.0 dependency and indirect dependencies
go.sum	Checksum entries for new dependencies
licenses/sigs.k8s.io/lws/LICENSE	Apache 2.0 license for LWS dependency
licenses/golang.org/x/sync/LICENSE	BSD license for golang.org/x/sync dependency
licenses/github.com/google/btree/LICENSE	Apache 2.0 license for btree dependency
licenses/github.com/fsnotify/fsnotify/LICENSE	BSD license for fsnotify dependency
docs/kthena/docs/reference/helm-chart-values.md	Minor formatting change (trailing whitespace)

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

pkg/model-serving-controller/controller/lws_controller.go

docs/proposal/modelserving-role-support-leaderworkerset.md

Copilot · 2026-01-06T08:05:00Z

pkg/model-serving-controller/controller/lws_controller.go

+		workerReplicas = 0
+	}
+
+	roleReplicas := int32(1)


Potential semantic mismatch between proposal and implementation. The proposal document (section 4.1, line 88) states that spec.replicas should map to ModelServing replicas, defining "the number of replicas for ServingGroup". However, the implementation sets roleReplicas to a hardcoded value of 1 (line 212) instead of using lws.Spec.Replicas. This means the LWS replicas are being used for ModelServing.Spec.Replicas (number of groups), while each role within a group always has exactly 1 replica. Verify if this mapping is correct according to the LWS semantics, as it may not align with user expectations from the LWS API.

pkg/model-serving-controller/controller/lws_controller.go

docs/kthena/docs/reference/helm-chart-values.md

pkg/model-serving-controller/controller/lws_controller.go

…oller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

pkg/model-serving-controller/controller/lws_controller.go

…oller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

charts/kthena/charts/workload/templates/kthena-controller-manager/rbac/cluster-role.yaml

docs/proposal/modelserving-role-support-leaderworkerset.md

hzxuzhonghu · 2026-01-14T03:30:12Z

go.mod

 	sigs.k8s.io/structured-merge-diff/v6 v6.3.0
 	sigs.k8s.io/yaml v1.6.0
-	volcano.sh/apis v0.0.0-20251222111824-40fed4793a6f
+	volcano.sh/apis v1.13.2-0.20260105020123-deb066235db8


@LiZhenCheng9527 Please help confirm whether the subgroup supported in in this release, otherwise should revert this

Support subgroup in this release

pkg/controller/controller.go

pkg/model-serving-controller/controller/lws_controller.go

hzxuzhonghu · 2026-01-14T03:41:43Z

pkg/model-serving-controller/controller/lws_controller.go

+	}
+
+	lwsInformerFactory.Start(ctx.Done())
+	kthenaInformerFactory.Start(ctx.Done())


I think start here maybe unsafe, since we have also started it in other controller. We should make sure all the resource event handler registered before informer started, otherwise we may miss some resource events

The NewSharedInformerFactory and start were split to ensure that all event handlers were registered before all informer started

hzxuzhonghu · 2026-01-14T03:42:45Z

@zhiweideren We should also add some test cases covering the translation, especially note the corner case

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

…odelServing Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

…lServing Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

zhiweideren · 2026-01-15T09:32:45Z

@zhiweideren We should also add some test cases covering the translation, especially note the corner case

done.add lws_controller_test.go

hzxuzhonghu

nit: now we create separate informers each controller, which could be shared.

pkg/model-serving-controller/controller/lws_controller.go

…ttings in LWS controller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

hzxuzhonghu · 2026-01-21T02:45:38Z

/lgtm

@zl-cheng @LiZhenCheng9527 @git-malu take another look?

…elsering_support_lws_issuer_407 # Conflicts: # charts/kthena/charts/workload/templates/kthena-controller-manager/rbac/cluster-role.yaml

volcano-sh-bot · 2026-01-22T09:06:39Z

Adding label do-not-merge/contains-merge-commits because PR contains merge commits, which are not allowed in this repository.
Use git rebase to reapply your commits on top of the target branch. Detailed instructions for doing so can be found here.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

hzxuzhonghu · 2026-01-22T09:49:04Z

/lgtm

hzxuzhonghu · 2026-01-22T09:59:26Z

/approve

hzxuzhonghu · 2026-01-22T09:59:30Z

/approve

volcano-sh-bot · 2026-01-22T09:59:36Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: hzxuzhonghu

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [hzxuzhonghu]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

volcano-sh-bot · 2026-01-22T09:59:39Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: hzxuzhonghu

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [hzxuzhonghu]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

volcano-sh-bot requested review from LiZhenCheng9527 and git-malu December 30, 2025 09:43

volcano-sh-bot added the size/L label Dec 30, 2025

gemini-code-assist bot reviewed Dec 30, 2025

View reviewed changes

docs/proposal/modelserving-role-support-leaderworkerset.md Outdated Show resolved Hide resolved

docs/proposal/modelserving-role-support-leaderworkerset.md Outdated Show resolved Hide resolved

zhiweideren force-pushed the modelsering_support_lws_issuer_407 branch from cc62948 to 276da2a Compare January 6, 2026 06:07

volcano-sh-bot added size/XL and removed size/L labels Jan 6, 2026

hzxuzhonghu requested a review from Copilot January 6, 2026 07:55

Copilot started reviewing on behalf of hzxuzhonghu January 6, 2026 07:56 View session

Copilot AI reviewed Jan 6, 2026

View reviewed changes

zhiweideren force-pushed the modelsering_support_lws_issuer_407 branch from 40f82b2 to 509fb04 Compare January 12, 2026 12:31

feat: add support for LeaderWorkerSet (LWS) API in ModelServing contr…

5afb248

…oller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

LiZhenCheng9527 reviewed Jan 13, 2026

View reviewed changes

pkg/model-serving-controller/controller/lws_controller.go Show resolved Hide resolved

feat: add support for LeaderWorkerSet (LWS) API in ModelServing contr…

48c6ff9

…oller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

zhiweideren force-pushed the modelsering_support_lws_issuer_407 branch from 509fb04 to 1a59635 Compare January 13, 2026 12:33

feat: add support for LeaderWorkerSet (LWS) API in ModelServing contr…

b6e3c59

…oller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

zhiweideren force-pushed the modelsering_support_lws_issuer_407 branch from 1a59635 to b6e3c59 Compare January 13, 2026 12:55

zhiweideren changed the title ~~[draft] ModelServing Role Support LeaderWorkerSet~~ ModelServing Role Support LeaderWorkerSet Jan 14, 2026

hzxuzhonghu requested a review from Copilot January 14, 2026 02:03

Copilot started reviewing on behalf of hzxuzhonghu January 14, 2026 02:03 View session

hzxuzhonghu reviewed Jan 14, 2026

View reviewed changes

Copilot AI reviewed Jan 14, 2026

View reviewed changes

zhiweideren added 2 commits January 15, 2026 15:50

refactor: improve LeaderWorkerSet compatibility and code clarity in M…

283947a

…odelServing Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

test: add unit tests for ModelServing construction from LeaderWorkerSet

788472e

Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

volcano-sh-bot added size/XXL and removed size/XL labels Jan 15, 2026

feat: implement LWS controller initialization and integration in Mode…

7c92e3e

…lServing Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

hzxuzhonghu previously approved these changes Jan 19, 2026

View reviewed changes

pkg/model-serving-controller/controller/lws_controller.go Outdated Show resolved Hide resolved

volcano-sh-bot added the approved label Jan 19, 2026

hzxuzhonghu reviewed Jan 19, 2026

View reviewed changes

pkg/model-serving-controller/controller/lws_controller.go Outdated Show resolved Hide resolved

hzxuzhonghu reviewed Jan 19, 2026

View reviewed changes

pkg/model-serving-controller/controller/lws_controller.go Show resolved Hide resolved

feat: add permissions for apiresources and adjust informer factory se…

cb1dc83

…ttings in LWS controller Signed-off-by: liuhaiyu <liuhaiyu13@huawei.com>

zhiweideren dismissed hzxuzhonghu’s stale review via cb1dc83 January 19, 2026 07:59

volcano-sh-bot removed the approved label Jan 19, 2026

zhiweideren requested a review from hzxuzhonghu January 19, 2026 08:21

volcano-sh-bot assigned hzxuzhonghu Jan 21, 2026

volcano-sh-bot added the lgtm label Jan 21, 2026

Merge branch 'main' of https://github.com/zhiweideren/kthena into mod…

65d96a6

…elsering_support_lws_issuer_407 # Conflicts: # charts/kthena/charts/workload/templates/kthena-controller-manager/rbac/cluster-role.yaml

volcano-sh-bot added do-not-merge/contains-merge-commits and removed lgtm labels Jan 22, 2026

volcano-sh-bot added the lgtm label Jan 22, 2026

volcano-sh-bot added the approved label Jan 22, 2026

volcano-sh-bot merged commit 0259a9e into volcano-sh:main Jan 22, 2026
14 of 16 checks passed

ModelServing Role Support LeaderWorkerSet #609

ModelServing Role Support LeaderWorkerSet #609

Uh oh!

Conversation

zhiweideren commented Dec 30, 2025

Uh oh!

gemini-code-assist bot commented Dec 30, 2025

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!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Key Changes

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 6, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hzxuzhonghu Jan 14, 2026

Choose a reason for hiding this comment

Uh oh!

LiZhenCheng9527 Jan 22, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

hzxuzhonghu Jan 14, 2026

Choose a reason for hiding this comment

Uh oh!

zhiweideren Jan 15, 2026

Choose a reason for hiding this comment

Uh oh!

hzxuzhonghu commented Jan 14, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

zhiweideren commented Jan 15, 2026

Uh oh!

hzxuzhonghu left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hzxuzhonghu commented Jan 21, 2026

Uh oh!

volcano-sh-bot commented Jan 22, 2026

Uh oh!

hzxuzhonghu commented Jan 22, 2026

Uh oh!

hzxuzhonghu commented Jan 22, 2026

Uh oh!

hzxuzhonghu commented Jan 22, 2026

Uh oh!

volcano-sh-bot commented Jan 22, 2026

Uh oh!

volcano-sh-bot commented Jan 22, 2026