---
cited_by:
  - '[[thoughts/inference/qwen3.8-serving#3-kv-cache--state-what-lives-where-per-request|qwen4next §3]]'
  - '[[thoughts/inference/qwen3.8-serving#4-pr-reading-order-vllm-main-git-show-sha|qwen4next §4]]'
date: '2026-11-11'
kind: pr
local:
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/core/kv_cache_manager.py
  - file:///Users/aarnphm/workspace/vllm/vllm/v1/kv_cache_interface.py
merged: '2026-09-14'
number: 55027
source:
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/distributed/kv_transfer/kv_connector/v1/mooncake/store/scheduler.py
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/distributed/kv_transfer/kv_connector/v1/mooncake/store/connector.py
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/v1/core/kv_cache_manager.py
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/v1/kv_cache_interface.py
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/distributed/kv_transfer/kv_connector/v1/mooncake/store/worker.py
  - https://github.com/vllm-project/vllm/blob/58d45fd767ff81229afd92543400f5c8c93f4000/vllm/distributed/kv_transfer/kv_connector/v1/mooncake/store/coordinator.py
state: merged
tags:
  - vllm
  - vllm-pr
  - new-architectures
title: '#55027 [Bugfix][KV Connector] MooncakeStore: exclude non-prefix-cacheable (QSA ring) groups; fix align-mode check'
topic: new-architectures
url: https://github.com/vllm-project/vllm/pull/55027
created: '2026-11-11'
modified: '2026-11-11'
published: '2026-11-11'
pageLayout: default
slug: hinterland/papers/maps/prs/55027
permalink: https://aarnphm.xyz/hinterland/papers/maps/prs/55027.md
generator:
  quartz: v4.6.0
  hostedProvider: Cloudflare
  baseUrl: aarnphm.xyz
full: https://aarnphm.xyz/llms-full.txt
---
Fix [#54726](https://github.com/aarnphm/aarnphm.github.io/issues/54726): Qwen3.8-Flash-Next with prefix caching and `MooncakeStoreConnector` fails because its QSA ring scratch cache is transferable, but is not valid hash-addressed prefix data. · full dump: [[thoughts/inference/vllm-prs/55027|#55027]]

