[CNCF LFX Proposal] Volcano/Kthena Build a KVCache-aware scheduler E2E test suite

### CNCF Project

Volcano

### Term

2026 Term 3 (Sep-Nov)

### Program Name

Build a KVCache-aware scheduler E2E test suite

### Program Description

## Description:
Kthena router already has a scheduler plugin E2E framework covering prefix-cache, least-request, least-latency, lora-affinity, random, and gpu-usage. The KVCache-aware scheduler plugin also exists, but lacks robust end-to-end coverage for the full runtime path involving the router, mock backends, tokenization, ZMQ events, and Redis.

This project aims to build a reusable KVCache-aware scheduler E2E test suite on top of the existing router plugin framework. In addition to validating routing behavior, the suite should provide reusable setup, cache-state injection, observation, and assertion utilities for future KVCache-aware test scenarios.

## Expected Outcome:

1. Build a reusable KVCache-aware E2E test suite under `test/e2e/router/`, including shared fixtures and helpers for llm-d-sim, the ZMQ bridge, Redis, cache-state injection, request generation, and routing assertions.
2. Validate the complete data path: dispatch KV-cache ownership events through ZMQ, verify the resulting Redis mappings, and confirm that the router uses them when scoring replicas.
3. Provide a focused command or Make target for running the suite independently, and document how contributors can add future KVCache-aware scenarios.
4. Integrate the suite into CI with deterministic cleanup and useful failure diagnostics for router, Redis, ZMQ, and backend state.

### Technologies

Go, Kubernetes, Kind-based E2E testing, Redis, ZMQ, observability/metrics, and distributed-system debugging.

### Skills same as Technologies?

- [x] Yes, the required skills are the same as the technologies listed above.

### Required/Desirable Skills

_No response_

### Mentors

Jinyu Zhou | @FAUST-BENCHOU | 2319109590@qq.com | benchou8
Jprakash | @katara-Jayprakash | katarajayprakash@icloud.com | lfxkatara1

### Upstream Issue URL

https://github.com/volcano-sh/kthena/issues/1328

### Application Prerequisites

- [x] Resume
- [x] Cover Letter
- [ ] School Enrollment Verification
- [ ] Participation Permission from school or employer
- [ ] Coding Challenge
- [ ] Custom Prerequisite (fill in details below)

### Coding Challenge URL

_No response_

### Custom Prerequisite Name

_No response_

### Custom Prerequisite Description

_No response_

### Custom Prerequisite — File Upload

- [ ] Yes — completion of this task requires the mentee to submit a file.

---
**LFX program:** [CNCF - Volcano: Build a KVCache-aware scheduler E2E test suite (2026 Term 3)](https://mentorship.lfx.linuxfoundation.org/project/16e5821d-2ae3-4145-a729-1b0c755b613b)

Start by reviewing the existing scheduler plugin E2E framework under test/e2e/router/ and the upstream issue linked in the proposal. The work is complete when reusable fixtures and helpers cover llm-d-sim, ZMQ, Redis, cache-state injection, requests, and routing assertions; the suite has an independent command, CI integration, cleanup, diagnostics, and contributor documentation.

Written by the indexing model from the issue text.