[TENT] Add UB failover, telemetry and runtime scheduling (TENT_3) - #48

Merged

zchuango merged 486 commits into

LinQuickDev:ub/tent3-integrationfrom

Sherlock0203:ub/tent3-integration

Sep 28, 2026

This pull request is big! We're only showing the most recent 250 commits

Commits on Sep 2, 2026

[TransferEngine] Keep RDMA owner CQ polled during active handshake (#3814)
alogfans
authored
04fd028
[TENT] Fix DeviceSelector inflight quota accounting (#3511)

authored
408b831

Commits on Sep 3, 2026

[PG] Fix CUDA enqueue-stream serialization (#3798)
caozhanhao
authored
69d41d5
[PG] Add support for torch==2.14.0 (#3842)
UNIDY2002
authored
886feb8
[Bugfix][TE] Use the only ADXL endpoint when dest has one engine (#3835)

authored
e4f552c
[Bugfix][TENT] Keep RailMonitor state across COW topology refreshes (#3832)

staryxchen
and
cursoragent
authored
6d946ef
[TENT]: Deduplicate TentMetrics singleton via shared tent_metrics library (#3797)

Colors-111
and
ruanzhao
authored
904b8ab
[TENT] Route HP TCP lanes across configured rails (#3834)
jacklin78911-collab
authored
7feb434
[Store] Preserve replica IDs during standby restore (#3811)

Icedcoco
and
Yuchen Kou
authored
65707bd
[Store] Enable fenced batch OpLog writer (#3810)

Icedcoco
and
Yuchen Kou
authored
3c84139
[Store] Structure ReplicaSelectionConfig environment settings (#3840)
bitborne
authored
6e87ff8
[Store] Align OffsetAllocatorBackendConfig with config layout (#3823)

bitborne
and
Aionw
authored
dd3ed28
fix(te): free wr_depth_list_ of zero-QP endpoints in ~RdmaEndPoint (#3846)

rishabhsinha17
and
claude
authored
fa75330
test: migrate placement, replication, and task scenarios to the DSL (#3702)

rishabhsinha17
and
claude
authored
3632717
[Store] Complete KV event publication (#3818)

authored
7e8ca11
[Store] Introduce placement index and replica allocator (#3704)
Aionw
authored
9e01833
[Store] Self-heal dangling LOCAL_DISK replicas in Client::Put (#3711)
he-yufeng
authored
8d063be
[P2P-Store] Unregister buffers on GetReplica failure paths (#3783)
chlins
authored
8c1edf4
[Bugfix] Fail SSD eviction tests on read errors (#3738)
migarci2
authored
d6e0960
[P2P-Store] Release batch IDs on all performTransfer exit paths (#3770)
chlins
authored
6957e76
[Build] conductor: build skeleton and common layer (#3825)

authored
5c68052
[TENT] Add native UB endpoint bootstrap and READ/WRITE data path (#3796)

authored
0f422b9
Bump google.golang.org/grpc in /mooncake-p2p-store/src/p2pstore (#3827)
dependabot[bot]
authored
e58ad95

Commits on Sep 4, 2026

Commits on Sep 5, 2026

[CI] Separate concurrent SSD write and read phases (#3870)
xiaono1
authored
c329f19

Commits on Sep 6, 2026

[Reshard] Add Store-backed model-weight snapshots (#3772)
Bo-Vincent
authored
cb9b380
[TransferEngine] Clear pending notify when freeing batch (#3891)

CorgiBoyG
and
guorongjie
authored
e9f91f7

Commits on Sep 7, 2026

[Bugfix][TENT] Predict deadline MLU from a metered wire rate and the bytes ahead (#3780)

SongOf
and
maxlisongsong
authored
4a91ce7
[Store] Support Dummy external buffer registration (#3721)
zxpdemonio
authored
9466cde
test: migrate storage-tier RPC scenarios to the DSL (#3847)

rishabhsinha17
and
claude
authored
345dc35
[CI/Build] Add Kylin support to dependencies installer (#3906)

LiChangXin-KylinSoft
and
LiChangXin-KylinSoft
authored
698543a
[CI/Build] Upload pre-release wheels to GitHub Releases (#3908)
stmatengss
authored
78566d4
[Store] Expose explicit soft pin action and TTL in C ABI and Go/Rust bindings (#3881)
CAICAIIs
authored
0fc9c00
[TENT] Disable Nagle on HP TCP client sockets (#3883)
jacklin78911-collab
authored
3a54ed4
[TENT] Pin dispatching_bytes_ to the eligible bytes in Dispatching (#3910)

SongOf
and
maxlisongsong
authored
17ea901
[TENT] Drain RDMA CQ and dest GPU before tearing down MRs (#3899)

staryxchen
and
cursoragent
authored
120714d
[Store] Add bounded standby promotion handoff (#3841)

Icedcoco
and
Yuchen Kou
authored
f30e10a
[Store] Structure TransferSubmitterConfig environment settings (#3850)
bitborne
authored
6e46302
[Store] Structure LocalFileSnapshotConfig environment settings (#3872)

bitborne
and
Aionw
authored
8215e2d
[Store] Structure RpcTimeoutConfig environment settings (#3871)

bitborne
and
Aionw
authored
8c0951b
[Store] Structure FilePerKeyConfig environment settings (#3853)
bitborne
authored
de85784
[Feature] Add per-file storage mode to storage_benchmark_v1 (#3454)
rayeaster
authored
bce9a7c
[Store] Stabilize HA under OpLog pressure (#3860)
tensorcopy
authored
1caf8f8

Commits on Sep 8, 2026

[CI/Build] Run current weight snapshot API tests in nightly (#3930)
migarci2
authored
4158f9d
[Bugfix][TransferEngine] Prevent peer failures from latching RDMA contexts inactive (#3559)

Dao007forever
and
claude
authored
715d1cf
[Store] Fail-stop supervisor on OpLog writer terminal failure (#3923)

Icedcoco
and
Yuchen Kou
authored
5d7f6c9
[CI/Build] Simplify CI orchestration and share smoke services (#3913)
Aionw
authored
f84c27a
[TENT] Add configuration lifecycle foundation (#3836)

bugkeep
and
jacklin78911-collab
authored
8715bb7
[EP] Add NCCL Device API backend for ElasticBuffer (#3449)
akhillanger
authored
d7f99ed
[TE] Add metadata versioning (#3837)
alogfans
authored
17ca605
[Bugfix][TENT] Serialize lanes sharing a queue pair on its slice queue (#3924)

SongOf
and
maxlisongsong
authored
95da554
[TENT] Fix HP TCP coalescing limits and short READ stalls (#3931)
jacklin78911-collab
authored
ac865da
[TENT] Say why the drop and arbitration MLU read different bytes_ahead (#3907)

SongOf
and
maxlisongsong
authored
82fff04
[Store] Migrate eviction and concurrency scenarios to the MasterScenario DSL (#3927)

rishabhsinha17
and
Aionw
authored
6678761
[Store] Split utility code by function (#3587)
Aionw
authored
a25719d
[Store] Route tensor _from writes through multi-buffers (#3722)

zxpdemonio
and
stmatengss
authored
3771d83
[Fix] TENT gid lid refresh (#3920)

TTThanos
and
TTThanos
authored
9d9088e
[TENT] Spin with pause before yielding in TicketLock (#3925)

SongOf
and
maxlisongsong
authored
2ca843e

Commits on Sep 9, 2026

[TE] Keep fabric mem Store-only; drop GRC fabric_memory enablement (#3955)
ascend-direct-dev
authored
23f3108
[TENT] Pin failover policy at transfer submission (#3934)
jacklin78911-collab
authored
d087726
[TENT] Skip idle GPUs when synchronizing dest CUDA contexts (#3945)

staryxchen
and
cursoragent
authored
0ba751f
[Store] Validate DummyClient registration of external pinned buffers (#3712)

zxpdemonio
and
stmatengss
authored
140e1f4
[Bugfix][TransferEngine] Export the requested range for dma-buf registration under CUDA VMM (fixes #2511) (#3538)

authored
e389a85
[Py] Fix batch_id leak on timeout and submit failure paths (#2350)
jfeng18
authored
42d9786
[Integration] Fail Python init when allocator support is unavailable (#2404)
he-yufeng
authored
4ed199e
TENT: deliver self-targeted notifications in-process (#3464)

xiaodouzi666
and
Frank
authored
fe92882
[Store] Introduce SegmentPool catalog and lifecycle (#3705)
Aionw
authored
f1ec715
[Store] Structure NvmeKvConnectorConfig environment settings (#3948)
bitborne
authored
8ffe105
[Store] Structure NoFRegisterConfig environment settings (#3949)

bitborne
and
Aionw
authored
feae277
[Store] Structure MmapArenaConfig environment settings (#3956)
bitborne
authored
70cec1e
[Store] Structure ClientNumaConfig environment settings (#3966)
bitborne
authored
740da2d
[Store] Structure CxlSegmentConfig environment settings (#3967)

bitborne
and
Aionw
authored
a6bfd17
[Store] Wire opt-in batch OpLog snapshots into standby runtime (#3950)

Icedcoco
and
Yuchen Kou
authored
3ab1a10
[Store] Add recoverable client liveness and asynchronous offboarding (#2991)

Doube-Suzerain
and
zhuyuening
authored
b905925

Commits on Sep 10, 2026

[Store] Sweep stale handles in bounded write-lock batches (#3593)
pjdurden
authored
f88d4c9
[TransferEngine] Add POSIX SHM transport for same-host DRAM copies (#3918)

staryxchen
and
cursoragent
authored
bf511f5
[TransferEngine] Prefer THP for shared-segment mmap without a HugeTLB pool (#3952)

ascend-direct-dev
and
cursoragent
authored
cbadcfd
[EP] Support NCCL ElasticBuffer recovery after rank replacement (#3963)
akhillanger
authored
a88bd08
[TENT] Export EGM host memory over multi-node NVLink (opt-in) (#3978)

xiaguan
and
claude
authored
6c07cb7
[Store] Release GIL during client setup (#3947)

LCAIZJ
and
leichao.lc
authored
d6da96a
[Reshard] Add KV cache reshard planning (#3564)
LLLL114
authored
6acb781
[CI/Build] Add self-hosted MI350X ROCm external PD tier (#3338)

andyluo7
and
Aionw
authored
eeaca79
[Bugfix] Preserve calling thread CPU affinity across SPDK env init (#3904)

NUABO
and
he-yufeng
authored
ef90acf
[Store] Add DFS read/write metrics and diagnostic logging (#3855)

authored
f6fcf19
[Store] Support online DFS shard capacity expansion (#3984)
zupengwang
authored
430c6e7
[P2PStore] Fix GetReplica metadata recheck retries (#3921)
waizuichougou
authored
5accaaf
[Store] Fail KV benchmarks when a worker raises (#3938)

030611
and
codex
authored
9b9e196
[TE] Report buffered JSONL write failures in tebench (#3992)

anranxia
and
jipengtian
authored
9cfb338

Commits on Sep 11, 2026

Commits on Sep 12, 2026

[Store] Support RealClient CUDA tensor reads (#3723)

zxpdemonio
and
stmatengss
authored
a8106d4
[TENT] Refresh TCP endpoints without replaying uncertain writes (#4006)
jacklin78911-collab
authored
2042b5b
[TENT] Refresh HP TCP endpoints after pre-send connection failures (#4007)
jacklin78911-collab
authored
40ba83b
TENT: derive notification QP attributes from endpoint params (#4064)

xiaodouzi666
and
xiaodouzi666
authored
6ae74b3

Commits on Sep 13, 2026

Commits on Sep 14, 2026

[Bugfix][TENT] Plan RDMA slices from the block size: no empty slices, and fold short tails (#3981)

SongOf
and
maxlisongsong
authored
c7aafbe
[TENT] Drain standard TCP workers during quiesce (#4005)
jacklin78911-collab
authored
1553a3b
[Store] Add a per-tenant eviction watermark so quota pressure evicts instead of rejecting (#3805)
gongwei-130
authored
662cd44
[Store] Structure RedisConnectionConfig environment settings (#4036)
bitborne
authored
d642e9f
[TENT] Add CUDA coverage for TCP data-path round trips (#4082)
CharlesXu-HQ
authored
1272940
[Bugfix][TENT] Keep engines alive for Python allocation guards (#4087)
jacklin78911-collab
authored
925a3ec
[TransferEngine] Enrich Ascend init and register-mem error logs with context (#4084)

Cheng-China
and
cursoragent
authored
db752df
perf(store): avoid redundant query result copies (#4085)
waizuichougou
authored
a3a2ad2
[TENT] Add Intel XPU platform to the Transfer Engine (#4030)
XinyuYe-Intel
authored
fa601df
[Bugfix][TENT] Fix Python send_probe under MC_USE_TENT (#4081)
CharlesXu-HQ
authored
97f302b
[Bugfix][TENT] Initialize byte counts when polling staged transfers (#4088)
jacklin78911-collab
authored
8a4f7f3
[Bugfix][TE] Fix 64KB alignment for default memory allocator (#4057)
gygdh-001
authored
22ef4d3
[TENT] Honor forced TCP in classic compatibility layer (#4027)

ToLiveAndLove
and
staryxchen
authored
83f9fac
[Bugfix][Store] Bind hugepage store segments onto agent-mode engines (#3917)
ascend-direct-dev
authored
5ac26e6
[Bugfix][Store] Set ACL context on DummyClient copy RPCs (#3998)

ascend-direct-dev
and
cursoragent
authored
38da705
[Store] Add OpLog writer runtime metrics (#4052)

Icedcoco
and
Yuchen Kou
authored
2a82e35
[Store] Rebootstrap standbys behind the compaction floor (#4048)

Icedcoco
and
Yuchen Kou
authored
77334f7
[Store] Add fork-safe SegmentPool snapshot codecs (#3706)
Aionw
authored
ec6e5e6
[Store] Drain in-flight RPCs and keep pools alive across client teardown (#3943)
he-yufeng
authored
fecdda7
[Conductor] conductor: tiered prefix indexer (#4041)

Misak2333
and
Chase-Rong
authored
7dee8db
[Bugfix][TENT] Forward classic RDMA device filters (#4028)

TTThanos
and
TTThanos
authored
6cc1a10

Commits on Sep 15, 2026

[Store] Shrink metadata maps after ClearStaleHandles sweeps (#4103)

Colors-111
and
ruanzhao
authored
cd6f10f
[PG] Fix P2P epoch matching and destroy/reinit test ordering (#4106)
caozhanhao
authored
f8fca5f
[Doc] Fix vLLM PD proxy guidance for the bootstrap-based MooncakeConnector (#4071)
CAICAIIs
authored
9288ee6
[TENT] Align Ascend Direct with async AutoConnect-only transfer. (#3933)
ascend-direct-dev
authored
fdc3eed
[Store] Add post-publish snapshot object GC (#4049)

Icedcoco
and
Yuchen Kou
authored
e0cdff8
[TransferEngine] Report Redis error replies as failures in the metadata plugin (#4107)
he-yufeng
authored
fad7261
[Bugfix][TENT] Detect dead peers in CheckSegmentStatus under MC_USE_TENT (#4109)

TTThanos
and
TTThanos
authored
fdf4553
[CI/Build] Fix scheduling race in SHM copy test (#4121)
jacklin78911-collab
authored
91b3042
perf(store): move batch write replica responses (#4102)
waizuichougou
authored
9fa8fdc
[Store] Add bounded batch OpLog deletion primitive (#4112)

Icedcoco
and
Yuchen Kou
authored
1d7a189
[Bugfix][Store] Properly enable O_DIRECT for OffsetAllocator vector I/O (#3615)

authored
cd71eee
[Store] Add object storage adapter and OSS integration (#3714)
fcczzz
authored
1a0c0a4
[CI/Build] Fix missing DeleteRange override in snapshot GC test (#4131)

anranxia
and
jipengtian
authored
a19d361
[TE] Fix transfer byte accounting races and completion caching (#3986)

anranxia
and
jipengtian
authored
9fb95ed

Commits on Sep 16, 2026

[Conductor] Add KV event protocol parsers (#4123)

authored
2459448
[Bugfix][TENT] Take the handler map lock when serving an RPC (#4147)

SongOf
and
maxlisongsong
authored
8442365
[Store] Centralize MasterService test access in a test-only peer (#4135)
Aionw
authored
a61a987
[TENT] Map compatibility API failures to classic error codes (#4139)

anranxia
and
jipengtian
authored
99a6bed
[Store] Structure OssAdapterConfig environment settings (#4146)
bitborne
authored
736987f
[Wheel] Migrate SSD administration to unified Python package (#3630)
Aionw
authored
38a1b69
fix: correct VMM teardown order in TENT MNNVL (#4108)
jfeng18
authored
74a929a
[Store] Make re-offload of persisted keys an idempotent skip in BatchOffload (#3135)
he-yufeng
authored
53d4132
[Store] Group NVMe KV implementation in a dedicated directory (#4134)
Aionw
authored
5df20cd
[CI/Build] Fix the USE_NOF build and NoF e2e on Ubuntu (#4079)
Ziy1-Tan
authored
f86289c
[TENT] Support priority and remote access options in the compatibility API (#4114)

anranxia
and
jipengtian
authored
97abb51
[Store] Extract per-object metadata and runtime state headers from MasterService (#4127)
CAICAIIs
authored
28644ce
[CI/Build] Isolate HA durable-finalize tests from tenant eviction (#4155)

Icedcoco
and
Yuchen Kou
authored
321b1f8
[Store] Freeze restored snapshot test eviction worker (#4156)

Icedcoco
and
Yuchen Kou
authored
211db0d
[BugFix][TENT] Keep first owner pending in ProgressWorker refill test (#4066)

staryxchen
and
cursoragent
authored
518781a
[Bugfix] Honor MC_FORCE_HCA before automatic GPU P2P fallback (#4151)
Alibaba-GuoYouzhi
authored
5ebe5fc
[Bugfix][TENT] Disable HP TCP when forcing TCP (#4113)
jacklin78911-collab
authored
ffe0135

Commits on Sep 17, 2026

[Store] Add durable awaiter to OrderedOpLogWriter (#4141)
0z5a
authored
455ca6b
[CI] Make the Go store integration link self-contained again (#4150)
he-yufeng
authored
c57745c
[Store] Add snapshot-validated batch OpLog pruning (#4142)

Icedcoco
and
Yuchen Kou
authored
2191fcc
[Bugfix][Store] Poll TENT batches with event completion enabled (#4143)
jacklin78911-collab
authored
93b88e4
[CI/Build] Add Hygon HCU PR validation (#4100)
Leewzh
authored
157c98a
[Store] Structure HaClusterNamespaceConfig environment settings (#4165)
bitborne
authored
0d63529
[Store] Structure FilereadWorkerPoolConfig environment settings (#4166)
bitborne
authored
b081523
[Store] Structure ClientHostIdentityConfig environment settings (#4119)
bitborne
authored
214d456
[Store] Structure S3ClientConfig environment settings (#4098)
bitborne
authored
f49a559
[Bugfix][Store] Register ShmHelper memory with SPDK for NoF zero-copy transfers (#3717)
NUABO
authored
b97af08
fix(tent): preserve cached handle after send probe (#4115)
ToLiveAndLove
authored
d783db4
[TENT] Fix first buffer address lookup in the compatibility API (#4138)

anranxia
and
jipengtian
authored
beca983
Feat/support store shm copy (#4110)
Luis-xu
authored
d9e4dcb
[Doc] Explain HP TCP multi-rail concurrency tuning (#4009)
jacklin78911-collab
authored
1a6c9bc
[TransferEngine] Cache the last-hit MR and index the buffer list for selectDevice (#4132)

staryxchen
and
cursoragent
authored
582b4b6
[TENT] Retry read-only segment metadata once after RPC failure (#4163)
Bo-Vincent
authored
c4d6328
[Store] Add batch snapshot and retention capacity metrics (#4173)

Icedcoco
and
Yuchen Kou
authored
5e73976
[CI/Build] Allow authorized fork checkout in Hygon CI (#4186)
Leewzh
authored
3e626f2

Commits on Sep 18, 2026

[Store] Add the per-object entry, object route and group table (#4161)
CAICAIIs
authored
12b344d
[Bugfix][Store] Support DFS replicas in ranged get sessions (#3903)
wp-bc
authored
90e6e24
[Doc] Nest Transfer Engine transport guides under Supported Protocols (#4206)

ykwd
and
cursoragent
authored
2a69368
[Bugfix][TENT] Keep a slice alive until the completion its post owes is handled (#4092)

SongOf
and
maxlisongsong
authored
2abe68b
[Store] Add SSD offload task lifecycle metrics (#4188)

authored
cab6c13
[Store] Expose client-observed Master heartbeat status and freshness (#4158)
WyattJia
authored
095c68c
[Store] Migrate lightweight Python utilities to unified package (#3586)
Aionw
authored
78465c5
[TENT]: Rate-limit TENT failure-storm logs & fix rail-cooldown defects (#3946)

Colors-111
and
ruanzhao
authored
1e09816
[TENT] Improve Ascend direct transport logging (#4201)
Cheng-China
authored
3362856
[TransferEngine] Wait classic tebench batches with getBatchTransferStatus (#4193)

staryxchen
and
cursoragent
authored
8aeecf7
[Wheel] Migrate Python CLI modules to unified package (#3626)
Aionw
authored
92cc759
[TENT] Keep a retiring notify QP published until its completions are drained (#4062)

xiaodouzi666
and
xiaodouzi666
authored
31dedeb
[CI/Build] Limit ROCm wheel builds to two parallel jobs (#4190)
Aionw
authored
24bfb98
[Store] Export process and allocator memory metrics, and complete the jemalloc build option (#4208)
DeyunYang
authored
d2785b0
[CI/Build] Extract shared steps for declarative PR and nightly workflows (#4175)
Aionw
authored
d9ae4df
[Store] add reusable ranged-read snapshots (#4095)

zxpdemonio
and
stmatengss
authored
2f13d40

Commits on Sep 19, 2026

[Store] Add weight metadata lifecycle core (#4116)

Bo-Vincent
and
stmatengss
authored
97030ca

Commits on Sep 20, 2026

[Store] Migrate async Python client to unified package (#3624)
Aionw
authored
eeb0512
[TENT] Reduce HP TCP READ receive overhead (#4205)
jacklin78911-collab
authored
8eeed7c
[TransferEngine]Add grouped scatter tasks for TCP (#4200)
LLLL114
authored
e6b2874
[TransferEngine] Cut WorkerPool submit locks with batched enqueue and rail count (#4189)

staryxchen
and
cursoragent
authored
675ca5c
[TENT] Add Intel XPU transport, routing and staging to the Transfer Engine (#4157)
XinyuYe-Intel
authored
003e0bb
[Doc] Streamline Quick Start (#4225)
ykwd
authored
32c020a
[Doc] Consolidate Mooncake Store API docs into one section (#4198)

ykwd
and
cursoragent
authored
24037ae
[Doc] Group design documents by module in the docs sidebar (#4196)

authored
9b5adcd
TENT: fall back to control-plane RPC when the RDMA notification channel is unavailable (#4068)

xiaodouzi666
and
xiaodouzi666
authored
68ce238

Commits on Sep 21, 2026

[Bugfix][Conductor] Clamp query match to the queried token count (#4164)

pjdurden
and
pjdurden
authored
251fc50
[Bugfix][Store] Isolate fileread worker count assertions from runtime threads (#4210)
Aionw
authored
c0678d6
[Store] Add ProbeKey/BatchProbeKey RPCs for lease-free existence probes (#3801)
duhaode520
authored
a816e5c
[Bugfix][Store] Make local MEMORY preference independent of replica order (#3662)

authored
847c8cc
[Store] Add etcd capacity soak and NOSPACE recovery operations (#4195)

Icedcoco
and
Yuchen Kou
authored
7faef38
[TE] Support ScaleFabric SHCA NICs (#3591)
huojianqiangg
authored
4d10d16
Merge branch 'kvcache-ai/main' into ub/tent3-integration
Sherlock0203
committed
d87191f
[Store] Parallelize OSS batches with independent request-scoped CURLM (#4105)
particle128
authored
fc07007
[Conductor] Add ZMQ event client with replay and reconnect (#4159)

authored
fe0d23e

Commits on Sep 22, 2026

[TENT] Bound UB bootstrap fan-out and keep rail rebuild telemetry
Sherlock0203
committed
2ecba9e
Merge branch 'kvcache-ai/main' into ub/tent3-integration
Sherlock0203
committed
5050e4c
[Bugfix][TENT] Allow default host allocations with SHM (#4250)
jacklin78911-collab
authored
328eb9d
[CI/Build] Publish ROCm nightly wheels to TestPyPI (#4266)
andyluo7
authored
a87818d
[TENT] Cover the RDMA context enable failure paths and link-speed events without an RNIC (#4263)

SongOf
and
maxlisongsong
authored
a6ea18c
[TransferEngine] Reuse last-hit device for consecutive slices in the same MR (#4259)

staryxchen
and
cursoragent
authored
4336273
[TransferEngine] Add opt-in NVIDIA Data Direct registration (#4267)
aurickq
authored
7364f79
[TransferEngine] Reuse selectTransport across consecutive same-target_id requests (#4273)

staryxchen
and
cursoragent
authored
b60c7d9
[TransferEngine] Add optional tebench submit/wait JSONL log (#4260)

staryxchen
and
cursoragent
authored
debb637
[TransferEngine] Add native RDMA notifications (#4214)

anranxia
and
jipengtian
authored
fd72779
[TENT] Detect Hygon hycu module for GPU-direct RDMA (#4276)
huojianqiangg
authored
2989ef0
[Store] Add immutable bucket allocator for DFS replicas (#4124)
wp-bc
authored
dbd46e6

Commits on Sep 23, 2026

[Doc][TENT] Add the deadline scheduling design doc (#4283)

SongOf
and
maxlisongsong
authored
0a0d5b4
[Bugfix][Store] Preserve tenant in BatchReplicaClear RPC (#4278)
XFDG
authored
8edde52
[TransferEngine] Reuse batch-local transport and single-device RDMA preparation (#4279)

authored
1016469
[CI] Publish explicit CUDA flavor tags (#4246)
tpiperatgod
authored
b696950
[Bugfix][TENT] Clear previous buffers when reading segment info (#4229)
jacklin78911-collab
authored
5600015
[TransferEngine] Add HOST_NUMA Fabric allocation core (#2966)

neverhook
and
wugou.cyf
authored
621e36b
[Store] Prevent object_already_exists log storm (#4285)
riZZZhik
authored
186f6ec
[Bugfix][TE] Register direct ACL VMM fabric memory as device (#4286)
csxxs
authored
5772765
Merge branch 'kvcache-ai/main' into ub/tent3-integration
Sherlock0203
committed
0c9410a