Repository navigation
feat(execution): settle native task spill through existing retention - #918
Draft
Haoming9527 wants to merge 1 commit into
Draft
Haoming9527 wants to merge 1 commit into
Haoming9527 wants to merge 1 commit into
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Refs #872. Stacked on #917 (
work/872-native-raw-spill, base691ec977dc832df9d182524ebbd0109866f94734). This is the task-output integration slice; #872 remains open.What was wrong
The native task queue retained bounded tails but did not configure the shared raw-spill store or hand complete native files to the existing retention ledger. That prevented the production fresh executor from requiring recoverable output parity. An unconfirmed capture join also needed to seal its owner before publishing the failure.
The change
Configure the allocation-only store before Windows kernel readiness and advertise
task-spill-v1alongsidetask-v1. Task readers retain the 64,000-byte tail, expose complete live prefixes up to 64 MiB, and withdraw the full path on storage/cap failure. Final settlement records only EOF-confirmed, error-free, sealed native streams after confirmed cleanup, with the actual kernel PID and captured Session ID, then notifies the existing retention scheduler. The original host execution path shares the settlement helper.Bound capture joins and retain poisoned ranges. Seal the owner before terminal publication and trigger independent transport shutdown even if publishing fails. The public host retires the owned kernel; late output cannot restore a successful cleanup claim.
The actual-process CI probe covers memory boundary, full stdout/stderr recovery through existing file permissions, live Jobs metadata, cap withdrawal, cancellation, retention and kernel crash cleanup. Bilingual docs and changelog are included. No dependency, version, model-facing schema or prompt change. Production fresh activation and persistent sessions remain subsequent integration work.
Validation
cargo test --locked --manifest-path native/Cargo.toml --bin dsh-ops-worker task::: 11 passed, 0 failed, 1 ignored; the ignored child fixture is explicitly invoked by its parent.cargo clippy --locked --manifest-path native/Cargo.toml --workspace --all-targets -- -D warnings,cargo fmt --all --check, worker build andnpm run build: passed.51ec139dc8e7dbb9fe7881ee71aa25c691e070bfeb122f50e7520a6bfe7383e6; retained report SHA-25629a0f9ac4320a09f66b719302c3024f4069a954de64758acc817e0f02d2938dc.git diff --check: passed.pull_requestCI run: https://github.com/T-Auto/dsh-ops/actions/runs/43490453643 — passed, including aggregateCI required, on final head454123e090636ae4a1ec861fa4faa6298e331e05.Environment
Windows x64, Node v26.12.1, Rust 1.99.0; local source worktree. Common stack originates from main
e724444416fed01eb1f75916ec556c0a286e73f9. The actual probe uses public host subprocess/sandbox/Jobs services with the internal task adapter; it does not substitute for the production fresh authority bridge or accepted persistent-runtime verification. No release or runtime pin was changed.