docs(design): update the P2 leftover section to the shipped/decided reality#9
Merged
Conversation
houseme
force-pushed
the
docs/design-p2-status
branch
from
b813cee to
269582d
Compare
…eality The "对 P2 主体实现的遗留项(本 spike 不覆盖)" section was written at spike completion (2026-07-07) and predates the #1102/#1144/#1145/#1159 implementation waves, so it now misdescribes the state. Verified each item against the code and rewrote the section with three explicit outcomes — implemented, closed-by-decision, or correctly-deferred: - eventfd reaping: implemented; AsyncFd-without-a-driver-thread closed (Drop can't await -> API break). - ring lifecycle: the singleton was replaced by per-disk sharding (probe_and_start_sharded); a singleton conflicts with per-disk isolation. - O_DIRECT read_at_direct: implemented, AND ecstore now wires it natively (rustfs/rustfs#4649) - the section still pointed at #4645's temporary route. - three read shapes: positioned pread_bytes wired; streaming reads are a measured NO-GO (backlog#1144), not open work. - per-disk probe cache + runtime errno latch: implemented (backlog#1101). - registered buffers / write path: still correctly deferred. Also correct the premise reference in "要证明的问题": the drop-in-flight scenario the crate exists for is realized on production GETs, and its primary trigger is the reader-setup quorum loop (create_bitrot_readers_until_quorum_ all_shards: FuturesUnordered io_primitives.rs:1362, break :1403, drop(reader_tasks) :1428), where data shards are read eagerly through UringBackend::pread_bytes -> driver.read_at().await; a task still parked at that await drops its ReadHandle in Submitted state and issues ASYNC_CANCEL. The old text pointed at a drifted line (:1455) and attributed the drop only to the decode-phase FuturesUnordered. Note in "这是什么" that P2 main has since shipped gray-off, fix the conclusion line to match, add the post-integration page-cache-reclaim contract (rustfs/rustfs#4662), and point the 5-item test matrix at the now-15 tests. The ownership model, invariants, and design constraints are unchanged. Co-Authored-By: heihutu <heihutu@gmail.com>
houseme
force-pushed
the
docs/design-p2-status
branch
from
269582d to
746e16a
Compare
houseme
added a commit
that referenced
this pull request
Jul 10, 2026
Supersedes the earlier "restore README" intent. The maintainer decision is to make 0.1.0 publish-ready but hold the actual crates.io release, so: - README: keep the crates.io/docs.rs badges (they go live on release), on top of #8's corrected formatting (de-indented O_DIRECT doctest, compact table) that #9's stale base had reverted. The Status note now says 0.1.0 is publish-ready with `publish = false` held until a maintainer cuts the release, rather than "not published". - Cargo.toml: keep `publish = false` but document that it is a hold, not a blocker — metadata is complete and every dependency is on crates.io, so the flip-and-`cargo publish` is a one-step maintainer action. - CHANGELOG: reframe the note the same way and record that the `[Unreleased]` section will move under `[0.1.0]` unchanged on release. No code change; no `cargo publish` performed (that stays a maintainer action with their token). README/CHANGELOG/Cargo.toml only. Co-Authored-By: heihutu <heihutu@gmail.com>
houseme
added a commit
that referenced
this pull request
Jul 10, 2026
Supersedes the earlier "restore README" intent. The maintainer decision is to make 0.1.0 publish-ready but hold the actual crates.io release, so: - README: keep the crates.io/docs.rs badges (they go live on release), on top of #8's corrected formatting (de-indented O_DIRECT doctest, compact table) that #9's stale base had reverted. The Status note now says 0.1.0 is publish-ready with `publish = false` held until a maintainer cuts the release, rather than "not published". - Cargo.toml: keep `publish = false` but document that it is a hold, not a blocker — metadata is complete and every dependency is on crates.io, so the flip-and-`cargo publish` is a one-step maintainer action. - CHANGELOG: reframe the note the same way and record that the `[Unreleased]` section will move under `[0.1.0]` unchanged on release. No code change; no `cargo publish` performed (that stays a maintainer action with their token). README/CHANGELOG/Cargo.toml only. Co-authored-by: heihutu <heihutu@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The
docs/DESIGN.mdsection 对 P2 主体实现的遗留项(本 spike 不覆盖) was written at spike completion (2026-07-07) and predates the #1102/#1144/#1145/#1159 implementation. It now misdescribes the state — most notably it still says ecstore O_DIRECT wiring is "remaining, see #4645's temporary route", when #4649 replaced that with native wiring.What I verified (each item, against the code)
probe_and_start_sharded)read_at_directpread_bytes✅ wired; streaming reads ⛔ NO-GO (#1144)Change
Rewrote the section with three explicit outcomes (implemented / closed-by-decision / deferred), updated the "这是什么" framing and the conclusion line (P2 main has since shipped, gray-off), added the post-integration page-cache-reclaim contract (#4662), and pointed the 5-item test matrix at the now-15 acceptance tests.
The ownership model, invariants, and test-matrix core are unchanged — they are the contract every shipped item still obeys. Docs-only.