v0.3.142026-11-03
This release brings major upgrades across Mooncake Store, Transfer Engine, TENT, Expert Parallelism (EP), Process Group (PG), ecosystem integrations, and hardware support, with 689 commits and 99 new contributors
Highlights
Transfer Engine & TENT
- Added POSIX shared-memory transport for efficient intra-node DRAM transfers.
- Improved transfer efficiency with pipelined operations, selective signaling, and native RDMA notifications, alongside optional adaptive RDMA congestion control.
- Enhanced TENT with asynchronous TCP transfers, multi-rail routing, large-read striping, load-aware lane selection, and idle connection reclamation.
- Expanded Prometheus metrics and improved RDMA connection lifecycle management, memory registration, and data integrity diagnostics.
Mooncake Store
- Added NVLink-based Extended GPU Memory (EGM) pooling on Grace-Blackwell and Vera Rubin platforms, enabling direct GPU access to shared host memory.
- Improved tenant-aware metadata management, quota accounting, and eviction with finer-grained locking and proactive reclamation.
- Enhanced client resilience with recoverable heartbeat states, asynchronous cleanup, and improved RPC isolation.
- Expanded external storage capabilities with DFS shard scaling, OSS/S3 backends, and optional SSD-to-DRAM prefetching.
Mooncake EP & Process Group (PG)
- Added an NCCL Device API backend for EP ElasticBuffer, supporting elastic peer management and rank replacement recovery.
- Introduced a new device collective framework in PG with IBGDA support.
- Added JIT compilation for EP CUDA backends and PG PyTorch adapters, removing PG's dependency on specific PyTorch versions.
Integration
- vLLM added cross-encoder embedding reuse through Mooncake Store for multimodal disaggregated serving.
- TokenSpeed integrated Mooncake Store as an L3 KV cache backend, alongside Transfer Engine support for PD/EPD disaggregation.
- TensorRT-LLM integrated Mooncake Store into KVCMv2 for content-addressed KV cache pooling and cross-instance prefix reuse.
- LMDeploy integrated Mooncake Store for shared KV cache pooling and cross-instance prefix reuse.
Platforms & Distribution
- Expanded hardware support across Hygon DCU, Cambricon MLU, MetaX MACA, Moore Threads MUSA, Biren SUPA, NVIDIA Data Direct, and AWS Trainium/EFA.
- Added Intel XPU support across Transfer Engine, Store, and EP, including PCIe P2P, GPUDirect RDMA, GDS, and host staging.
- Enhanced Ascend NPU support with the new MemFabric transport and mooncake-transfer-engine-memfabric wheel package, alongside native UB Transport, HCCS, and A5 RoCE capabilities.
- Expanded PyPI wheel support to Python 3.14.
- Added build support for macOS, Kylin OS, and RISC-V.
What's Changed
* [TENT] Remove global lock serialization on per-batch hot paths by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3648
* [TENT] Preserve per-request RDMA selection policy by @waizuichougou in https://github.com/kvcache-ai/Mooncake/pull/3647
* [Bugfix][TE] Restore fork state on memory unregister to prevent VMA exhaustion (#3639) by @XFDG in https://github.com/kvcache-ai/Mooncake/pull/3660
* [Build] Propagate YLT includes to PG device compiler by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3691
* [TENT] Guard RDMA worker latency samples behind latency logging flag by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/3690
* [CI/Build] Gate pre-releases through TestPyPI by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3672
* [Docs] Add FlagCX transport README by @MC952-arch in https://github.com/kvcache-ai/Mooncake/pull/3695
* [Bugfix][PG] Balance CUDA graph capture-depth accounting by @waizuichougou in https://github.com/kvcache-ai/Mooncake/pull/3701
* [Bugfix] [TransferEngine] Fence stale RDMA handshakes after QP reconstruction by @ehuohz in https://github.com/kvcache-ai/Mooncake/pull/3584
* TENT: flush a peer's pooled RPC connections when one of them fails by @xiaodouzi666 in https://github.com/kvcache-ai/Mooncake/pull/3684
* [TransferEngine] Reuse TCP CUDA staging buffers across chunks by @thunguo in https://github.com/kvcache-ai/Mooncake/pull/3562
* [TransferEngine] Log batchTransferSync failure paths by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3676
* [Bugfix][TE] Evict stale endpoints on RDMA port recovery by @XFDG in https://github.com/kvcache-ai/Mooncake/pull/3604
* [TENT] Cache metrics label cells to remove hot-path lock contention by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3677
* [TENT] Fix IPC relocation to not switch to the peer's device ordinal by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3678
* [TE] Bound RDMA batch MR registration by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3664
* [Store] Migrate object lifecycle and tenancy MasterService tests to the scenario DSL by @rishabhsinha17 in https://github.com/kvcache-ai/Mooncake/pull/3656
* [Doc] Add ROCm Wheel Badge and Simplify the README by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/3719
* [Bugfix][Store] Handle BatchEvict OpLog failures by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3422
* [EP] Migrate Python layer to unified package by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3635
* [Bugfix][TENT] Guard deferred batch liveness checks with the batch shard by @waizuichougou in https://github.com/kvcache-ai/Mooncake/pull/3725
* [Reshard] Bind logical transfer plans to runtime snapshots by @Bo-Vincent in https://github.com/kvcache-ai/Mooncake/pull/3634
* [Store] Decouple object routing from grouped-object semantics by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/3589
* [Store] Gate HA serving on successful standby restore by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3497
* [Store] Add batch OpLog snapshot bootstrap by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3642
* [EP] retain tensors for deferred receive hooks by @UNIDY2002 in https://github.com/kvcache-ai/Mooncake/pull/3732
* [Doc] Add ROCm package to Quick Start by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3720
* [Store] Bound eventual PutStart retries in scenario tests by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3748
* [TransferEngine][EP] Support Kimi K3 & guard against duplicate glog initialization by @kyx1999 in https://github.com/kvcache-ai/Mooncake/pull/3727
* [CI/Build] Add RISC-V build support by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/3590
* Revert "[CI/Build] Add RISC-V build support" by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/3759
* [TENT] Derive per-NIC bandwidth from the negotiated port speed by @SongOf in https://github.com/kvcache-ai/Mooncake/pull/3644
* [Bugfix][Store] Preserve selected local disk replica by @yyaoling in https://github.com/kvcache-ai/Mooncake/pull/3620
* [TENT] Share the CUDA IPC handle across buffers in one cudaMalloc segment by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3679
* [TENT] Recover from synchronous submit failures via transport failover by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3682
[TE] Shard TCP transport I/O across a pool of iocontext threads by @anranxia in https://github.com/kvcache-ai/Mooncake/pull/3724
* [TENT] Keep the worker from blocking on its own submit queue by @he-yufeng in https://github.com/kvcache-ai/Mooncake/pull/3683
* Fix while send error reset connection by @XiaokunDing in https://github.com/kvcache-ai/Mooncake/pull/3475
* tent: make Redis/Http metastore thread-safe by @XiaokunDing in https://github.com/kvcache-ai/Mooncake/pull/3470
* [Bugfix][Store] Remove partial files after short pwritev by @waizuichougou in https://github.com/kvcache-ai/Mooncake/pull/3766
* [Bugfix][TENT] Retry RDMA bootstrap after stale peer endpoints by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3755
* [Store] Add fenced snaps
v0.3.14-pre-4预发布2026-11-02
Pre-release built from main at commit 7d4670b67152cab643caa9a81e70bca39c782e8e.
- Git tag: v0.3.14-pre-4
- Python package version: 0.3.14rc4
- Core wheel matrix: CUDA 12, CUDA 13, and non-CUDA; x86_64 and aarch64; Python 3.10–3.14.
- NPU wheels: x86_64 and aarch64; Python 3.9–3.13, published as available by the existing best-effort workflow.
Build and publication are in progress. Wheels will be attached by the Pre-Release workflow. TestPyPI publication, installation smoke tests, and T-one integration have not yet completed.
Changes since the previous pre-release: https://github.com/kvcache-ai/Mooncake/compare/v0.3.14-pre-3...v0.3.14-pre-4
v0.3.14-pre-3预发布2026-10-30
v0.3.14-pre-2预发布2026-10-30
v0.3.14-pre预发布2026-10-26
Testing prerelease for Mooncake v0.3.14.
- Source commit: b46254dea9515ea6cc05fe4ec220545835582edd (main at tagging time).
- Git tag: v0.3.14-pre.
- Python package version: 0.3.14rc0 (PEP 440 normalization of the tag).
- The Pre-Release workflow builds CUDA 12, CUDA 13, non-CUDA, and NPU wheels for x86_64 and aarch64, attaches them here, and publishes to TestPyPI. It then runs indexed-wheel smoke tests and T-one integration validation.
- Wheel artifacts and test results are pending at publication time. This prerelease is intended for testing and is not a stable release.
Pre-Release workflow · Changes since v0.3.13.post1
v0.3.13.post12026-08-31
What's Changed
* [CI/Build] Keep 0.3.13 non-CUDA wheels CUDA-free by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3746
* [CI/Build] Backport TestPyPI pre-release gate to v0.3.13 by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3697
Full Changelog: https://github.com/kvcache-ai/Mooncake/compare/v0.3.13...v0.3.13.post1
v0.3.132026-08-26
What's Changed
[TENT] Fix iouring stale status in getTransferStatus by @jfeng18 in https://github.com/kvcache-ai/Mooncake/pull/2379
* ci: add manual "Cancel Queued and Running CI" workflow by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/3117
* ci: fix cancel-ci listing that matched no runs by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/3123
[TENT] Fix reqmap_ leak for single-task batches in ascend transport by @jfeng18 in https://github.com/kvcache-ai/Mooncake/pull/3116
* [PG] Forward collectives through the backend shim for PyTorch 2.13 by @mmangkad in https://github.com/kvcache-ai/Mooncake/pull/3122
* [Store] implement RFC #2756 batch-record HA OpLog writer by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/2826
Fix: allocationstrategy not taking effect when enable_ha is enabled on master by @gitgaoqian in https://github.com/kvcache-ai/Mooncake/pull/3077
* [CI] Add nightly build/release and test workflow by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/2950
* [CI] Add staryxchen for Mooncake TE's codeowner by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/3132
* [TransferEngine] Harden socket handshake frame handling by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/3050
* [Bugfix] Fix double free of RDMA slice on device-selection failure by @SongOf in https://github.com/kvcache-ai/Mooncake/pull/3125
* [tent] Fix Prometheus histogram silent-drop and clean stale config keys by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3126
* [PG][2/N] Introduce Control Plane by @caozhanhao in https://github.com/kvcache-ai/Mooncake/pull/2455
fix(rdma): auto-chunk MRs larger than device maxmr_size (#2017) by @jiejingzhangamd in https://github.com/kvcache-ai/Mooncake/pull/2644
* [P2P-Store] Fix double-close panic, refCount copy, and offset reset by @jfeng18 in https://github.com/kvcache-ai/Mooncake/pull/2375
* [CI/Build] master image: install ibverbs-providers and publish a CUDA 13 flavor by @tpiperatgod in https://github.com/kvcache-ai/Mooncake/pull/3121
[CI/Build] Remove CUDA dependency from mooncakemaster by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3119
* [CI/Build] Add ARM64 Non-CUDA release wheels by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3096
* [Store] Add shared thread-local random helpers by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2954
* [Store] Fix segment capacity metric leak on master service teardown by @liuzijing2014 in https://github.com/kvcache-ai/Mooncake/pull/3154
* [TENT] Harden ShmTransport backing-file and mapping lifecycle by @alexps9 in https://github.com/kvcache-ai/Mooncake/pull/3145
* [Bugfix] Fix double free of UB/Barex slices on device-selection failure by @he-yufeng in https://github.com/kvcache-ai/Mooncake/pull/3146
* [Store] Surface partial replica allocation in PutStart and document allocation semantics by @liangxu2000 in https://github.com/kvcache-ai/Mooncake/pull/3156
* [Store] Fix SSD total capacity showing incorrectly after master restart by @NUABO in https://github.com/kvcache-ai/Mooncake/pull/3069
* [Store] Fix null-ptr crash in PrepareStorageBackend by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/3138
[Store] Add dlopen Rust backend and WITHSTORECSHARED shared library by @donghun-furiosa in https://github.com/kvcache-ai/Mooncake/pull/3078
* [Bugfix][Common] Avoid etcd keepalive pings without active streams by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3144
* [TransferEngine] Add NCCL host and device transport backend by @akhillanger in https://github.com/kvcache-ai/Mooncake/pull/2852
[Store] Make RPC client I/O threads configurable via MCRPCCLIENTIO_THREADS by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3004
* [Misc] Update CODEOWNERS for docs, CI, and HA by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/3175
* [CI/Build] Align Asio with yalantinglibs target by @CanYangGetYang in https://github.com/kvcache-ai/Mooncake/pull/3170
* [tent] Support mixed DRAM+VRAM segments and multi-transport allocation in tebench by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/3127
* [CI] Fix nightly workflow: drop removed build-with-ep input by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/3172
* [TE] fix(efa): bind a CUDA context before registering GPU memory by @whn09 in https://github.com/kvcache-ai/Mooncake/pull/3177
[Bugfix][TE] Respect remoteaccessible in RDMA verbs MR registration by @RuiqingFeng in https://github.com/kvcache-ai/Mooncake/pull/3151
* [Store] Scope capacity metric release to the serving MasterService by @liuzijing2014 in https://github.com/kvcache-ai/Mooncake/pull/3168
* [Store] Fix NoF heartbeat config silently ignored in HA mode by @NUABO in https://github.com/kvcache-ai/Mooncake/pull/3149
* [Doc] Document master KV event publisher flags by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3150
* [Dep] bind wildcard listen address (0.0.0.0/::) without going through getaddrinfo by @pjdurden in https://github.com/kvcache-ai/Mooncake/pull/2919
* [TENT] Fix failover attempt metrics attribution by @anranxia in https://github.com/kvcache-ai/Mooncake/pull/3109
* [CI/Build] Deduplicate EFA and store CI workflows by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3163
* [CI/Build] Format only changed C++ lines by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3027
* [Bugfix][Store] Support JsonCpp header layouts across Linux distributions by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3193
* [Store] support tensor APIs with dummy client by @zxpdemonio in https://github.com/kvcache-ai/Mooncake/pull/2998
* [Store] Fix mem storage display showing 16777216 TB when no segment i… by @leonzzhu in https://github.com/kvcache-ai/Mooncake/pull/3080
* [Store] Add Optional Object Checksum Diagnostics by @yokinoshitayoki in https://github.com/kvcache-ai/Mooncake/pull/3003
[Store] Fix EraseMetadata leaking stale entries in offloadingobjects by @huniu20 in https://github.com/kvcache-ai/Mooncake/pull/3160
* [Bugfix][TE] Split RDMA slices at MR boundaries and handle submit errors by @RuiqingFeng in https://github.com/kvcache-ai/Mooncake/pull/3198
* feat(rdma): enable PCI relaxed ordering by default by @1998zxn in https://github.com/kvcache-ai/Mooncake/pull/3205
* [TransferEngine] Reject peer segment descriptors with more keys than devices by @eharris128 in https://github.com/kvcache-ai/Mooncake/pull/3209
* [PG] Fix single-rank initialization and yalantinglibs dependency by @caozhanhao in https://github.com/kvcache-ai/Mooncake/pull/3164
* [CI/Build] Unify CI and release wheel builds by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3137
* [CI] Fix TONE SGLang Qwen3.5 cache gating by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3195
* [Doc] Update RBG integration example to workloads.x-k8s.io/v1alpha2 by @tpiperatgod in https://github.com/kvcache-ai/Mooncake/pull/3224
* Bump google.golang.org/grpc from 1.79.3 to 1.82.1 in /mooncake-common/etcd by @dependabot[bot] in https://github.com/kvcache-ai/Mooncake/pull/3235
* [Store] Add exponential backoff for ordered oplog writer retries by @Icedcoco in https://github.com/kvcache-ai/Mooncake/pull/3166
* [ROCm] Add ROCm/HIP wheel build, CI, and release parity with CUDA (#3171) by @andyluo7 in https://github.com/kvcache-ai/Mooncake/pull/3184
* [Common/TE] Replace hardcoded CUDA paths with CUDAToolkit discovery by @caozhanhao in https://github.com/kvcache-ai/Mooncake/pull/3215
* [TENT] Reject malformed numeric metrics environment overrides by @Hubert-Zhu in https://github.com/kvcache-ai/Mooncake/pull/3228
* [TE] perf(efa): allow bounding the batch MR registration fan-out by @whn09 in https://github.com/kvcache-ai/Mooncake/pull/3210
* [Store] Consolidate string parsing primitives by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/3179
* [Bug]: [SSD] DSV4 enable SSD of
v0.3.12.post12026-07-25
What's Changed
* [build] Strip shared libraries to reduce NPU wheel size by @JieTang66 in https://github.com/kvcache-ai/Mooncake/pull/2202
* [Docs] Update README with citation details by @UNIDY2002 in https://github.com/kvcache-ai/Mooncake/pull/2190
* [Doc]: Update vLLM LMCache guide for MP interface by @fcczzz in https://github.com/kvcache-ai/Mooncake/pull/2209
* [Store] fix gauge overflow by separating DFS unlimited flag by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2152
[Store]: avoid INVALIDREPLICA error on empty offload by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2151
* [Store] Fix: Auto-recovery for SSD Offload after Master Restart by @Colors-111 in https://github.com/kvcache-ai/Mooncake/pull/2077
* [Store] Fix: validate HA backend availability during config parsing by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2111
* [Doc] Add missing aws-logo by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/2211
* [codex] Update snapshot object store docs by @Dao007forever in https://github.com/kvcache-ai/Mooncake/pull/2148
[TE] Python api "registermemory" & "batchregistermemory" support location param by @A-Liuhao in https://github.com/kvcache-ai/Mooncake/pull/2191
* [Bugfix][Store] Fix snapshot failure when OpLog has never been written by @leonzzhu in https://github.com/kvcache-ai/Mooncake/pull/2144
* [Build] Fix compile warnings across multiple components by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2193
* [Store] fix batch tensor allocation error code by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2226
* Fix MACA nvlink allocator build by mapping CUmemAllocationHandleType by @muma378 in https://github.com/kvcache-ai/Mooncake/pull/2227
* [Docs]: fix docs config, remove autodoc2, archive zh docs, add .agents/… by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2218
* [TransferEngine][MACA] Add missing CUDA-like type aliases for MACA compatibility by @Dayuxiaoshui in https://github.com/kvcache-ai/Mooncake/pull/2230
* [TENT] Add rule-based transport and device selection by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2079
* [TE] Add ProgressWorker skeleton by @ZhenyuePan in https://github.com/kvcache-ai/Mooncake/pull/2199
* [Store] remove invalid kMaxSliceSize assertion in AllocateBatch by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2165
* [CI] Add docs-check job to validate Sphinx build with -W by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2229
* [Store] Add opt-in grouped object routing semantics by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2180
* feat(store): add SPDK NoF worker pool by @Enigmo-x in https://github.com/kvcache-ai/Mooncake/pull/2172
[Store] (CI runtestswithssd failed / promotion-on-hit failed) eliminate race conditions by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2235
[TransferEngine][docs] document FIEFAUSEDEVICE_RDMA=0 for same-host EFA loopback by @Chelseatr in https://github.com/kvcache-ai/Mooncake/pull/2222
* [Doc] update overall picture by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/2249
* [CI] feat: pre-release ci workflow by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/2212
* [Doc] Clarify TENT failover poll behavior by @ZhenyuePan in https://github.com/kvcache-ai/Mooncake/pull/2208
[Security] Fix Go vulnerabilities in libetcdwrapper.so by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2250
* Build tent by @Dao007forever in https://github.com/kvcache-ai/Mooncake/pull/2089
* [Store] Clarify cache stats semantics by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2248
* feat(store): route NoF replicas through put and get by @Enigmo-x in https://github.com/kvcache-ai/Mooncake/pull/2247
* [Doc] Split LMCache vLLM MP and non-MP guides by @fcczzz in https://github.com/kvcache-ai/Mooncake/pull/2268
* [PG][EP] Fix engine.so runtime dependency by @mmangkad in https://github.com/kvcache-ai/Mooncake/pull/2255
* [Store] Implement tenant metadata map isolation by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2232
[EFA] Add MCEFACQTHREADS env var to cap CQ poller threads by @yuhuiaws in https://github.com/kvcache-ai/Mooncake/pull/2113
* [TransferEngine][ROCm] Add HIP dmabuf MR registration for AMD GPUs (fixes #751) by @andyluo7 in https://github.com/kvcache-ai/Mooncake/pull/2225
* [Store] enables the Ubtransport for Mooncake Store And optimize UrmaEndpoint by @zchuango in https://github.com/kvcache-ai/Mooncake/pull/2196
[Store] L2->L1 promotion-on-hit: observability metrics + maxper_heartbeat knob by @yzhan1 in https://github.com/kvcache-ai/Mooncake/pull/2176
* fix(metrics): show actual client-reported SSD capacity instead of infinite by @Oxygen56 in https://github.com/kvcache-ai/Mooncake/pull/2278
* [TE][Store] Fix IPv6 address parsing in connection endpoints by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2184
Fix P2PHANDSHAKE in dual-NIC container setups via MCRDMABINDADDRESS by @stmatengss with @Copilot in https://github.com/kvcache-ai/Mooncake/pull/2280
* [TE] IntraNode NVLink transport: update cuMemcpyAsync to BatchAsyc for CUDA version >= 12.8 by @TTThanos in https://github.com/kvcache-ai/Mooncake/pull/2251
* fix(wheel): exclude libfabric/libefa from auditwheel bundle to avoid dual-libfabric EFA conflict by @whn09 in https://github.com/kvcache-ai/Mooncake/pull/2271
* remote redis dependency by @jinke446 in https://github.com/kvcache-ai/Mooncake/pull/2109
* [Store] Robustify ConfigDict size parsing by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2206
fix: unify default clusternamespace to match master's DEFAULT_CLUSTE… by @leonzzhu in https://github.com/kvcache-ai/Mooncake/pull/2244
* [TE] fix: pass default port when parsing TENT RDMA bind a… by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2289
[Store] RemoveAll not deleting SSD offload files, enable storagebackend_->RemoveAll() in Client::RemoveAll by @Colors-111 in https://github.com/kvcache-ai/Mooncake/pull/2283
* [Store] Propagate tenant identity through object RPCs by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2288
* [Doc] add vLLM scenario-based landing pages and archive legacy docs by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2262
* [Store] Support tenant-aware async storage tasks by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2294
* [Build] Disable debug symbols (-g) in default compilation flags by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2285
* [TE] Fix TCP connection pool SIGSEGV by deferring cleanup with asio::post by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2174
* Store Extract 3FS logic into DistributedStorageBackend by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2234
* [TENT] Add policy name binding to transport selector by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2295
* [Doc] reorganize API reference with Python/C++/HTTP sub-indices by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2263
* [Store] Support SSD offload configuration in standalone store service by @ertcmm in https://github.com/kvcache-ai/Mooncake/pull/2261
[TransferEngine] Make TCP transport slice size configurable via MCTCPSLICESIZE by @gogongxt in https://github.com/kvcache-ai/Mooncake/pull/2308
fix(transferengine): improve auto gid selection and retry by @Bo-Vincent in https://github.com/kvcache-ai/Mooncake/pull/2269
[TE/TENT] Allow building tebench without USETENT by @00fish0 in https://github.com/kvcache-ai/Mooncake/pull/2322
* fix(ci): move sccache --show-stats to after build steps by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2303
* [Docs][1/N] Refactor Readme: update readme top link and badges by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/2304
* [Docs] [2/N] Refactor Readme: merges the Showcase and Components
v0.3.122026-07-23
What's Changed
docs: reflect WITHNVIDIA_PEERMEM change from CMake flag to runtime env var by @stmatengss with @Copilot in https://github.com/kvcache-ai/Mooncake/pull/2164
* feat(rdma): mlx5dv QP path diversity via UDP sport and LAG port balance by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/2175
* [TENT] Enhanced QoS and Slice Spraying for TENT by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2048
* [TENT] Add codeowner to the TENT directory by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2183
* Add release-npu workflow by @JieTang66 in https://github.com/kvcache-ai/Mooncake/pull/2178
[chore] Set default for WITHNVIDIA_PEERMEM to true by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/2192
* [Store] Add structured object store helper by @zxpdemonio in https://github.com/kvcache-ai/Mooncake/pull/2140
* Bump version to 0.3.11.post1 in pyproject.toml by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/2194
* [build] Strip shared libraries to reduce NPU wheel size by @JieTang66 in https://github.com/kvcache-ai/Mooncake/pull/2202
* [Docs] Update README with citation details by @UNIDY2002 in https://github.com/kvcache-ai/Mooncake/pull/2190
* [Doc]: Update vLLM LMCache guide for MP interface by @fcczzz in https://github.com/kvcache-ai/Mooncake/pull/2209
* [Store] fix gauge overflow by separating DFS unlimited flag by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2152
[Store]: avoid INVALIDREPLICA error on empty offload by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2151
* [Store] Fix: Auto-recovery for SSD Offload after Master Restart by @Colors-111 in https://github.com/kvcache-ai/Mooncake/pull/2077
* [Store] Fix: validate HA backend availability during config parsing by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2111
* [Doc] Add missing aws-logo by @ykwd in https://github.com/kvcache-ai/Mooncake/pull/2211
* [codex] Update snapshot object store docs by @Dao007forever in https://github.com/kvcache-ai/Mooncake/pull/2148
[TE] Python api "registermemory" & "batchregistermemory" support location param by @A-Liuhao in https://github.com/kvcache-ai/Mooncake/pull/2191
* [Bugfix][Store] Fix snapshot failure when OpLog has never been written by @leonzzhu in https://github.com/kvcache-ai/Mooncake/pull/2144
* [Build] Fix compile warnings across multiple components by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2193
* [Store] fix batch tensor allocation error code by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2226
* Fix MACA nvlink allocator build by mapping CUmemAllocationHandleType by @muma378 in https://github.com/kvcache-ai/Mooncake/pull/2227
* [Docs]: fix docs config, remove autodoc2, archive zh docs, add .agents/… by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2218
* [TransferEngine][MACA] Add missing CUDA-like type aliases for MACA compatibility by @Dayuxiaoshui in https://github.com/kvcache-ai/Mooncake/pull/2230
* [TENT] Add rule-based transport and device selection by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2079
* [TE] Add ProgressWorker skeleton by @ZhenyuePan in https://github.com/kvcache-ai/Mooncake/pull/2199
* [Store] remove invalid kMaxSliceSize assertion in AllocateBatch by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2165
* [CI] Add docs-check job to validate Sphinx build with -W by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2229
* [Store] Add opt-in grouped object routing semantics by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2180
* feat(store): add SPDK NoF worker pool by @Enigmo-x in https://github.com/kvcache-ai/Mooncake/pull/2172
[Store] (CI runtestswithssd failed / promotion-on-hit failed) eliminate race conditions by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2235
[TransferEngine][docs] document FIEFAUSEDEVICE_RDMA=0 for same-host EFA loopback by @Chelseatr in https://github.com/kvcache-ai/Mooncake/pull/2222
* [Doc] update overall picture by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/2249
* [CI] feat: pre-release ci workflow by @stmatengss in https://github.com/kvcache-ai/Mooncake/pull/2212
* [Doc] Clarify TENT failover poll behavior by @ZhenyuePan in https://github.com/kvcache-ai/Mooncake/pull/2208
[Security] Fix Go vulnerabilities in libetcdwrapper.so by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2250
* Build tent by @Dao007forever in https://github.com/kvcache-ai/Mooncake/pull/2089
* [Store] Clarify cache stats semantics by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2248
* feat(store): route NoF replicas through put and get by @Enigmo-x in https://github.com/kvcache-ai/Mooncake/pull/2247
* [Doc] Split LMCache vLLM MP and non-MP guides by @fcczzz in https://github.com/kvcache-ai/Mooncake/pull/2268
* [PG][EP] Fix engine.so runtime dependency by @mmangkad in https://github.com/kvcache-ai/Mooncake/pull/2255
* [Store] Implement tenant metadata map isolation by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2232
[EFA] Add MCEFACQTHREADS env var to cap CQ poller threads by @yuhuiaws in https://github.com/kvcache-ai/Mooncake/pull/2113
* [TransferEngine][ROCm] Add HIP dmabuf MR registration for AMD GPUs (fixes #751) by @andyluo7 in https://github.com/kvcache-ai/Mooncake/pull/2225
* [Store] enables the Ubtransport for Mooncake Store And optimize UrmaEndpoint by @zchuango in https://github.com/kvcache-ai/Mooncake/pull/2196
[Store] L2->L1 promotion-on-hit: observability metrics + maxper_heartbeat knob by @yzhan1 in https://github.com/kvcache-ai/Mooncake/pull/2176
* fix(metrics): show actual client-reported SSD capacity instead of infinite by @Oxygen56 in https://github.com/kvcache-ai/Mooncake/pull/2278
* [TE][Store] Fix IPv6 address parsing in connection endpoints by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2184
Fix P2PHANDSHAKE in dual-NIC container setups via MCRDMABINDADDRESS by @stmatengss with @Copilot in https://github.com/kvcache-ai/Mooncake/pull/2280
* [TE] IntraNode NVLink transport: update cuMemcpyAsync to BatchAsyc for CUDA version >= 12.8 by @TTThanos in https://github.com/kvcache-ai/Mooncake/pull/2251
* fix(wheel): exclude libfabric/libefa from auditwheel bundle to avoid dual-libfabric EFA conflict by @whn09 in https://github.com/kvcache-ai/Mooncake/pull/2271
* remote redis dependency by @jinke446 in https://github.com/kvcache-ai/Mooncake/pull/2109
* [Store] Robustify ConfigDict size parsing by @CAICAIIs in https://github.com/kvcache-ai/Mooncake/pull/2206
fix: unify default clusternamespace to match master's DEFAULT_CLUSTE… by @leonzzhu in https://github.com/kvcache-ai/Mooncake/pull/2244
* [TE] fix: pass default port when parsing TENT RDMA bind a… by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2289
[Store] RemoveAll not deleting SSD offload files, enable storagebackend_->RemoveAll() in Client::RemoveAll by @Colors-111 in https://github.com/kvcache-ai/Mooncake/pull/2283
* [Store] Propagate tenant identity through object RPCs by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2288
* [Doc] add vLLM scenario-based landing pages and archive legacy docs by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2262
* [Store] Support tenant-aware async storage tasks by @Lin-z-w in https://github.com/kvcache-ai/Mooncake/pull/2294
* [Build] Disable debug symbols (-g) in default compilation flags by @Aionw in https://github.com/kvcache-ai/Mooncake/pull/2285
* [TE] Fix TCP connection pool SIGSEGV by deferring cleanup with asio::post by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2174
* Store Extract 3FS logic into DistributedStorageBackend by @LujhCoconut in https://github.com/kvcache-ai/Mooncake/pull/2234
* [TENT] Add policy name binding to transport selector by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/22
v0.3.11.post12026-05-23
What's Changed
docs: reflect WITHNVIDIA_PEERMEM change from CMake flag to runtime env var by @Copilot in https://github.com/kvcache-ai/Mooncake/pull/2164
* feat(rdma): mlx5dv QP path diversity via UDP sport and LAG port balance by @staryxchen in https://github.com/kvcache-ai/Mooncake/pull/2175
* [TENT] Enhanced QoS and Slice Spraying for TENT by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2048
* [TENT] Add codeowner to the TENT directory by @alogfans in https://github.com/kvcache-ai/Mooncake/pull/2183
* Add release-npu workflow by @JieTang66 in https://github.com/kvcache-ai/Mooncake/pull/2178
[chore] Set default for WITHNVIDIA_PEERMEM to true by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/2192
* [Store] Add structured object store helper by @zxpdemonio in https://github.com/kvcache-ai/Mooncake/pull/2140
* Bump version to 0.3.11.post1 in pyproject.toml by @ShangmingCai in https://github.com/kvcache-ai/Mooncake/pull/2194
New Contributors
* @JieTang66 made their first contribution in https://github.com/kvcache-ai/Mooncake/pull/2178
Full Changelog: https://github.com/kvcache-ai/Mooncake/compare/v0.3.11...v0.3.11.post1