release: v0.8.3 — connection plane at scale (c10k/c1M campaign, idle conns −94%)

CHANGELOG.md

@@ -6,6 +6,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.8.3] — 2026-07-29
+
 ### Added
 - **Task-exit parking for idle connections (c1M P1, `--conn-park-secs`,
   default 60 s).** A plain-TCP monoio connection that stays idle past the
@@ -24,7 +26,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   downshift), and the tokio runtime. `--conn-park-secs 0` disables.
 
 ### Fixed
-- **c10k connection-plane wave (PR #TBD)** — from the 2026-07-29 empirical
+- **c10k connection-plane wave (PR #421)** — from the 2026-07-29 empirical
   review (`tmp/C10K-REVIEW.md`: 10k/25k live-connection ramps, idle-CPU
   measurement, symbolized flames, pipeline-ratchet A/B on the Linux VM):
   - **Pipeline memory ratchet fixed (W1).** One 1024-deep pipeline

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "moon"
-version = "0.8.2"
+version = "0.8.3"
 edition = "2024"
 rust-version = "1.94"
 description = "A high-performance Redis-compatible server written in Rust"

README.md

@@ -372,7 +372,8 @@ Redis 8.6.1.*
 | **v0.7.x** | Replication GA for multi-shard masters — `WAIT`/`ACK` across all six planes, 24 h kill-9 soak (zero acked-write loss) | **GA** |
 | **v0.8.0** | One Storage Kernel — kill-9-lossless on every plane (46-cell crash matrix) + 10× RAM datasets under disk offload | **GA** |
 | **v0.8.1** | Deploy-safe busy-poll (O3 contention governor auto-gates on shared cores) + single-shard tuning preset | **GA** |
-| **v0.8.2** (current) | Storage kernel unification — one value codec / spill pipeline / eviction entry point / accessor skeleton, millisecond-TTL fidelity fix, batched sync-eviction manifest fsyncs | **GA** |
+| **v0.8.2** | Storage kernel unification — one value codec / spill pipeline / eviction entry point / accessor skeleton, millisecond-TTL fidelity fix, batched sync-eviction manifest fsyncs | **GA** |
+| **v0.8.3** (current) | Connection plane at scale — idle conns 56.5 → 3.25 KB (−94%, task-exit parking + idle downshift + TLS diet), loud maxclients, striped registry, c1M-ready (1 M idle ≈ 3.3 GB) | **GA** |
 | **v0.9** | Horizontal scale — cluster-on-monoio + multi-shard replicas | planned |
 | **v1.0** | Every [`PRODUCTION-CONTRACT.md`](docs/PRODUCTION-CONTRACT.md) GA box ticked | gate |
 

RELEASES.md

@@ -1,5 +1,10 @@
 # Releases
 
+## v0.8.3 — 2026-07-29
+milestones: none (patch release: the c10k/c1M connection-plane campaign, rounds 1–5)
+waivers: none new — `XSHARD-READ-01` remains the open GA gap (ROADMAP R4), unchanged from v0.8.2.
+evidence: Patch release rolling up the connection-plane campaign (#421 rounds 1–4, #422 round 5) plus an SDK reconnect fix. **Headline: idle-connection memory 56.5 → 3.25 KB/conn (−94%); 1 M idle connections ≈ 3.3 GB.** The campaign, measured end-to-end in `tmp/C10K-REVIEW.md` and designed in `.planning/rfcs/c1m-connection-plane.md`: **(1) memory** — pipeline ratchet fixed (one 1024-deep pipeline permanently pinned a conn at ~217 KB; now 47 KB, W1), per-conn task future dieted 9.3 → 5.9 KB by boxing cold suspend points (P3), W11 two-stage idle park sheds buffers to a 512 B probe (43.5 → 19.9 KB, loss-free cancel-and-await on both drivers), TLS diet vendors monoio-rustls + monoio-io-wrapper to make the wrapper's 32 KB eager buffers lazy + released on idle (87.5 → 47.4 KB/TLS-conn, P4b), and task-exit parking (`--conn-park-secs`, default 60 s) ends the handler task entirely for long-idle plain-TCP conns, leaving a tiny readiness watcher (19.8 → **3.25 KB/conn**, P1 — no vendor patch; monoio's public `readable()` is race-free on both drivers); **(2) correctness/operability** — loud `-ERR max number of clients reached` + RLIMIT_NOFILE check (W3), 16-way striped client registry (W5), deadline-heap blocking sweep O(due) (W6), SPSC drain rotation (W7), affinity-funnel load gate (W8), loud MOON_URING bypass warning (T3). **Validation:** every wave A/B-measured on the Linux VM at 10 k conns (same-binary flag A/Bs for W11/P4b/P1); GCE t2a hardware proof — idle 33.1 → 17.9 KB/conn (−46%), pinned p=1/p=16 perf-neutral; wire-parity suites on one continuous connection across park/wake cycles green on kqueue AND io_uring (`tests/parked_idle_parity.rs`, `tests/tls_idle_downshift_parity.rs`, `tests/maxclients_reject_parity.rs`); parked conns stay in CLIENT LIST and are killable (kill's `shutdown(2)` wakes the watcher). Independent lifecycle review clean; one documented Low (transient CLIENT LIST invisibility during the data-wake respawn window, same observable as a reconnect race). This release touches no crash/persistence path; the crash-matrix + soak gate is dispatched on the RC to hold the ritual (soak-first-then-tag).
+
 ## v0.8.2 — 2026-07-20
 milestones: none (patch release: the storage-unification campaign W1–W6 + db.rs directory split)
 waivers: none new — `XSHARD-READ-01` remains the open GA gap (ROADMAP R4), unchanged from v0.8.1.