Skip to content

Releases: NoKV-Lab/NoKV

Release list

v1.0.0-alpha.1

v1.0.0-alpha.1 Pre-release
Pre-release

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 08 Jun 02:41
v1.0.0-alpha.1
cee155a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

NoKV is now a single-node, agent-native copy-on-write filesystem — "git for live, large, derived state." This first alpha consolidates the pivot from a distributed KV/DFS into a focused CoW workspace runtime for AI agents and ML pipelines.

Where git versions small text at dev time, NoKV gives snapshot / clone / diff / rollback over live filesystem namespaces (code + data + model weights) at runtime speed — O(metadata), not O(data) — and at scale.

Highlights

Copy-on-write workspace operations

  • snapshot — zero-copy MVCC version pin of a subtree (read-only, GC-protected).
  • clone — writable fork; shares object blocks zero-copy (same generation → same object keys); diverges on write (CoW); O(metadata). Spin up 100 agent workspaces off one base; each writes its own delta over shared blocks.
  • diff — Added / Removed / Modified between two subtrees; unchanged shared files are skipped.
  • rollback — revert a workspace to a prior snapshot via clone-from-snapshot + atomic graft (keeps inode identity); discarded delta becomes GC-reclaimable, restored blocks survive.
  • CLI: nokv clone, nokv diff (snapshot / rollback exposed at the service/RPC layer).

See docs/cow-workspaces.md.

Architecture

  • Single-node metadata engine (Holt 0.5.4) — local hot path, no quorum, no Redis/TiKV round-trip.
  • Object data on S3-compatible storage (RustFS / S3 via OpenDAL); 4 MiB block / 64 MiB chunk.
  • Structural zero-copy object layout: blocks/{mount}/{inode}/{generation}/{chunk}/{block} — forks share blocks by construction.
  • Durability ordering: object-first, metadata-atomic-after → a crash leaves GC-able orphan objects, never a dangling pointer.
  • The distributed/Raft metadata layer was removed in favor of a focused single-writer-owner model (read-only snapshots are shared by many readers).

Correctness & performance

  • Write path: ~213 MiB/s sequential (release build), durable — close drains to S3.
  • Six write-path correctness bugs root-caused and fixed (overlapping staged blocks from FUSE re-dispatch, manifest bloat past the metadata value limit, silent zero-fill on getattr, stale dentry CAS on mid-write setattr, body-wipe after chmod, xattr rejection).
  • GC safety — the owns_block_object_key invariant: an inode reclaims only blocks it minted; retained snapshot pins protect still-shared base blocks.
  • 322 tests passing; clippy -D warnings clean.

Scope & boundaries

  • Single-node: one writer-owner per workspace; many readers share snapshots.
  • POSIX via FUSE (Unix-only).
  • Not yet in this alpha: cloud sharding + owner leases, derivation lineage, branch/merge, a FUSE control surface for clone.

Distribution

GitHub release only — not published to crates.io. No crate version was changed.

Loading

v0.9.0

v0.9.0 Latest
Latest

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 14 May 18:01
v0.9.0
893c527

Highlights

  • Added the native fsmeta service path for distributed metadata workloads, including rooted mount lifecycle, WatchSubtree, SnapshotSubtree, quota fences, hard links, subtree authority handoff, fsmeta contract tests, and Docker/benchmark harnesses.
  • Introduced and hardened the Peras metadata fast path: compiler proof hardening, isolated benchmark coverage, operation-id scoping, read-path materialization, and release-blocking CLI startup safety for authority key loading.
  • Strengthened the distributed storage substrate with Eunomia authority grants, dynamic coordinator duty grants, rooted store membership, scan retry budgets, follower read modes, dependency-aware async apply, batched 2PC/RPC paths, and WAL/raftlog durability improvements.
  • Expanded release and evaluation surfaces: GHCR image publishing, Docker Compose demo hardening, official fsmeta workload profiles, fsmeta median/long benchmark CI, DCO enforcement, VitePress docs, repository code contract, TLA artifacts, and correctness/chaos smoke coverage.

Install

Binary archives are attached for Linux and macOS on amd64/arm64. Each archive includes nokv, nokv-config, nokv-fsmeta, fsmeta diagnostic tools, README.md, LICENSE, and raft_config.example.json.

shasum -a 256 -c checksums.txt

Container images are published as:

ghcr.io/feichai0017/nokv:v0.9.0
ghcr.io/feichai0017/nokv:0.9.0
ghcr.io/feichai0017/nokv:0.9
ghcr.io/feichai0017/nokv:latest

Docker Hub mirrors use the same tags when Docker Hub publishing is available:

docker.io/feichai0017/nokv:v0.9.0
docker.io/feichai0017/nokv:0.9.0
docker.io/feichai0017/nokv:0.9
docker.io/feichai0017/nokv:latest

What's Changed

  • feat(fsmeta): add native metadata service, fsmeta executor, server/client split, benchmark workloads, WatchSubtree, SnapshotSubtree, rooted mounts, quota fences, hard links, and subtree authority handoff.
  • feat(eunomia): add dynamic coordinator duty grants and authority protocol hardening.
  • feat(raftstore): add follower read modes, dependency-aware async apply, batched proposals, bounded scan retry, and 2PC rerouting/cleanup hardening.
  • feat(wal/raftlog): enforce durability and retention contracts, batch fsync durability requests, rebuild segment catalogs, and shard raft WAL managers.
  • perf(fsmeta): optimize placement, key shape routing, read paths, atomic mutate fast paths, TSO coalescing, and fsmeta benchmark paths.
  • perf(lsm): shard data-plane WAL managers, add shard hint cache, reduce read-path allocations, and improve write/read/compaction paths.
  • docs: replace mdBook with VitePress, add NoKV code contract, refresh public architecture/fsmeta docs, and document Peras architecture boundaries.
  • bench: add official fsmeta workload profiles, fsmeta benchmark profile bundles, Docker benchmark CI, and improved benchmark diagnostics.
  • ci: add GHCR Docker image publishing, DCO enforcement, CodeQL/Codecov stabilization, and updated GitHub Actions dependencies.
  • fix(root): defer grant key loading until authority use so release binaries can run read-only CLI commands without grant key environment variables.

Additional Changes

This release includes many cleanup and hardening commits across control-plane protocol vocabulary, package boundaries, error taxonomy, fsmeta recovery, coordinator liveness, meta-root timing, Docker demo reliability, test deflaking, generated artifacts, and security/dependency updates.

New Contributors

No first-time contributors in this release.

Full Changelog: feichai0017/NoKV@v0.8.0...v0.9.0

Assets 7
Loading

v0.8.0

v0.8.0

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 16 Apr 14:11
v0.8.0
dc05ec6

Highlights

  • Added the namespace listing research line, including a hardened listing contract, benchmark tooling, and paper-oriented framing for direct-children / stable pagination style namespace access.
  • Expanded the metadata-root and control-plane substrate with separated deployment, transport-backed replication, freshness / catch-up semantics, allocator fencing, and clearer rooted publish lifecycle boundaries.
  • Reorganized NoKV as a more maintainable distributed storage research platform: single-node engine code now lives under engine/*, DB/runtime boundaries were tightened, benchmark outputs were unified, and the top-level docs now explicitly describe the repository as an extensible research base.

What's Changed

Additional Changes

New Contributors

No first-time contributors in this release.

Full Changelog: feichai0017/NoKV@v0.7.2...v0.8.0

Loading

v0.7.2

v0.7.2

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 30 Mar 01:03
v0.7.2
73d64a5

Highlights

  • Hardened the single-node write path around a bounded MPSC commit queue, including correctness fixes for publish/wake behavior, a long-lived consumer session, batch draining, and sync-pipeline cleanup.
  • Tightened distributed truth-source and runtime boundaries: local region truth now stays apply/bootstrap-driven, peer region metadata is explicitly a local mirror, and scheduler-side PD publishing is kept separate from local catalog updates.
  • Clarified client routing/degraded behavior with explicit route-availability errors, unified retry entrypoints for single-key and multi-key flows, lazy store dialing, and cleaner transport/retry boundaries.
  • Aligned top-level architecture and runtime documentation with the current code paths.

What's Changed

Additional Changes

New Contributors

No first-time contributors in this release.

Full Changelog: feichai0017/NoKV@v0.7.1...v0.7.2

Loading

v0.7.1

v0.7.1

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 23 Mar 12:27
v0.7.1
71c0c0d

Highlights

  • Added practical LSM range-filter pruning for point reads and bounded scans, including exact-candidate point fast paths, table-internal block-range pruning, and read-path cleanup.
  • Tightened value separation and value-log defaults: CF/keyspace-aware routing landed, ordinary multi-bucket vlog routing remains enabled, ValueThreshold moved to 2048, and HotRing is now narrowed to an optional write-hot detector/throttling tool instead of a default data-path strategy.
  • Continued correctness and boundary cleanup across startup recovery, WAL/LSM integration, iterator semantics, range tombstones, region metadata APIs, and stale internal wrappers.

What's Changed

Additional Changes

New Contributors

Full Changelog: feichai0017/NoKV@v0.7.0...v0.7.1

Contributors

sreekar2307, dependabot, and 7 other contributors
Loading

v0.7.0

v0.7.0

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 03 Mar 05:02
v0.7.0
fbed3a2

Highlights

  • Added PD-based control plane integration for distributed mode (routing, metadata persistence, allocator bootstrap).
  • Removed standalone/local transaction APIs from the embedded path; transaction semantics are now aligned with distributed Percolator flow.
  • Continued iterator and LSM correctness improvements.

What's Changed

Additional Changes

New Contributors

Full Changelog: feichai0017/NoKV@v0.6.1...v0.7.0

Contributors

CyberSleeper, nothiny, and 2 other contributors
Loading

v0.6.1

v0.6.1

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 25 Feb 17:05
v0.6.1
7b01914

What's Changed

New Contributors

Full Changelog: feichai0017/NoKV@v0.6.0...v0.6.1

Contributors

vasilytrofimchuk, dependabot, and 4 other contributors
Loading

v0.6.0

v0.6.0

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 16 Feb 01:31
v0.6.0
557793e

What's Changed

New Contributors

Full Changelog: feichai0017/NoKV@v0.5.0...v0.6.0

Contributors

dependabot, zqr10159, and 3 other contributors
Loading

Release v0.5.0

Release v0.5.0

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 07 Feb 16:36
v0.5.0
3b9fc5a

v0.5.0

Highlights

  • Bucketed ValueLog (hashkv) with hot/cold routing for better large-value isolation and GC control
  • HotRing externalized and split into read/write tracking; new write-hot metrics exposed
  • TOML-based engine config loader for embedded deployments

Storage / ValueLog

  • Bucketed ValueLog with hot/cold buckets
  • Parallel VLog GC with compaction/backlog pressure control
  • Fixes for GC skip thresholds and candidate filtering

HotRing & Observability

  • Read/write HotRing split to avoid read hotness impacting write throttling
  • Added HotWriteKeys / HotWriteRing metrics
  • HotRing defaults and config normalization tuned

Config & UX

  • New TOML loader: LoadOptionsFile / ApplyOptionsFile
  • Example config: nokv.options.toml

Fixes

  • Hardened raftstore reads and MVCC scans
  • Iterator ordering and write-throttle alignment fixes

Docs

  • Expanded HotRing/VLog/Compaction docs
  • DBDB badge
Loading

Release v0.4.2

Release v0.4.2

Choose a tag to compare

View all tags
@feichai0017 feichai0017 released this 22 Jan 14:04
v0.4.2
ca19adb

This release provides critical security fixes, cache stability improvements, and enhanced CI/CD integration. It addresses issues identified by CodeQL and optimizes the W-TinyLFU cache implementation for long-term stability.

🚀 Key Highlights

🛡️ Security Fixes

  • LSM/FID Boundary Check: Fixed a high-severity integer overflow vulnerability in SSTable file ID parsing (utils.FID). This prevents potential memory corruption or logical errors when processing large file IDs. (Addressed CodeQL scan alert).

⚡ Cache Optimization (W-TinyLFU)

  • Automatic Freshness (Reset Mechanism): Fixed a bug where the threshold for Frequency Sketch and Bloom Filter resets was uninitialized. Resets now trigger every 10 * size operations, preventing historical data from polluting the frequency estimates.
  • Improved Entry Validation: Added explicit bounds checks in the Set path for the internal cache. Key hashes exceeding uint32 are now handled safely without silent truncation, ensuring consistency with the Get logic.

🛠 Infrastructure & CI/CD

  • AI-Powered Code Review: Integrated CodeRabbit for automated PR reviews, configured specifically for the Go toolchain (go-vet, golangci-lint).
  • Dependency Updates: Bumped github.com/panjf2000/ants/v2 to v2.11.4 for improved concurrency pooling stability.

📈 What's Changed

  • fix(cache): initialize reset threshold for TinyLFU by @feichai0017
  • fix(security): prevent integer overflow in FID parsing by @feichai0017
  • chore: add .coderabbit.yaml for intelligent PR analysis
  • deps: upgrade ants/v2 pool manager

🔗 Full Changelog
For a detailed list of all 100 commits since v0.4.1, please refer to the commit history
(feichai0017/NoKV@v0.4.1...v0.4.2).

Contributors

feichai0017
Loading