Issues

• [good-first-issue] distributed: convert Mooncake transfer-engine logging to %-format

  good first issuehelp wanted
  LMCache/LMCache#5731 · Filter by author maobaolong opened on Oct 14, 2026
  2

• [good-first-issue] distributed: convert Bigtable L2 adapter logging to %-format

  good first issuehelp wanted
  LMCache/LMCache#5730 · Filter by author maobaolong opened on Oct 14, 2026
  1

  

• [new feature]: Add lmcache-layerwise plugin

  LMCache/LMCache#5729 · Filter by author zxue2 opened on Oct 14, 2026
  2

• [Test][MP] Add unit tests for SerdeL2AdapterWrapper failure and shutdown paths

  LMCache/LMCache#5718 · Filter by author alany85 opened on Oct 14, 2026

• [Bug][CLI] query engine starts TTFT at the first answer token on the chat endpoint

  LMCache/LMCache#5715 · Filter by author theminjunchoi opened on Oct 14, 2026

• [Bug][MP/P2P] L2 attach timeouts leave untracked adapters after retries and peer removal

  LMCache/LMCache#5698 · Filter by author JasperChan929 opened on Oct 13, 2026

• [Bug][MP] CPU KV events withdraw a chunk while another L1 or salt copy is still live

  LMCache/LMCache#5694 · Filter by author theminjunchoi opened on Oct 13, 2026

• [Bug][MP] A DAX L2 adapter can map a device L1 already uses when set at startup

  LMCache/LMCache#5691 · Filter by author theminjunchoi opened on Oct 13, 2026

• [Bug] Worker-node enumeration races with controller registrations

  LMCache/LMCache#5678 · Filter by author AaravMehta-07 opened on Oct 12, 2026

• [BUG] Hybrid Mamba/GDN + MTP5: mixed local+external prefix hit with ≥2 locally-continued chunks ⇒ deterministic spec-decode degeneration (acceptance exactly 1.00, per-position 0.000×5)

  LMCache/LMCache#5675 · Filter by author zhangpeng871018 opened on Oct 12, 2026

• [Bug] --enable-segmented-prefix never returns a tail: the prefix prefetch trims at the first gap

  LMCache/LMCache#5670 · Filter by author shawnguyen-tensormesh opened on Oct 12, 2026

• [Bug] L1 read locks are anonymous: a TTL expiry silently drops one reader's locks and lets another reader release them

  LMCache/LMCache#5669 · Filter by author shawnguyen-tensormesh opened on Oct 12, 2026
  1

• [Bug][MP] Lookup read locks stay held when a request is aborted before its retrieve

  LMCache/LMCache#5656 · Filter by author theminjunchoi opened on Oct 11, 2026

• [Bug][MP] A failed retrieve reports vLLM's prefix-cache blocks as load errors

  LMCache/LMCache#5653 · Filter by author theminjunchoi opened on Oct 11, 2026

• [Bug] L2 benchmark loses timeout accounting after warmup filtering

  LMCache/LMCache#5638 · Filter by author mohan-zeyu opened on Oct 11, 2026

• [Bug] First-rank MLA retrieval stages the whole cache hit on the device

  LMCache/LMCache#5632 · Filter by author orthur2 opened on Oct 11, 2026

• [Bug] Delayed lookup cleanup can unpin a replacement after targeted clear

  LMCache/LMCache#5629 · Filter by author fieldwork1f1010-rgb opened on Oct 10, 2026

• [Bug] Bitmap OR truncation leaves padding bits visible to highest_set_bit()

  LMCache/LMCache#5628 · Filter by author fieldwork1f1010-rgb opened on Oct 10, 2026

• [Bug][MP] The coordinator keeps L1 placements for MP servers that are gone

  LMCache/LMCache#5626 · Filter by author theminjunchoi opened on Oct 10, 2026

• [Bug] lmcache bench engine rejects workloads without a max-output-length parameter when the length is unset

  LMCache/LMCache#5623 · Filter by author zhongerdeajie opened on Oct 10, 2026

• [Bug] cache-simulator cannot parse dtypes from its own server's lookup logs (torch-qualified dtype names)

  LMCache/LMCache#5619 · Filter by author zhongerdeajie opened on Oct 10, 2026

• [Bug][MP] Incompatible KV caches collide when FP8 and BF16 instances share Mooncake

  LMCache/LMCache#5616 · Filter by author wjinxu opened on Oct 10, 2026
  1

• [Bug][Operator] Some valid CR settings leave the engine or coordinator unable to start

  LMCache/LMCache#5607 · Filter by author theminjunchoi opened on Oct 10, 2026

• [Bug] CacheBlend on align-mode Mamba hybrids: a chunk without a state snapshot is never fingerprinted, so most documents cannot be blended

  LMCache/LMCache#5601 · Filter by author shawnguyen-tensormesh opened on Oct 10, 2026

• [Bug] Mamba hybrids without --separate-object-groups: prefix hits restore a state vLLM never computed

  LMCache/LMCache#5600 · Filter by author shawnguyen-tensormesh opened on Oct 10, 2026