fix(runtime): make unknown-job sweep tests deterministic #663

Merged

toasterson merged 1 commit from fix/runtime-sweep-deterministic-clock into main 2026-10-16 10:24:30 +00:00

Conversation 0 Commits 1 Files changed 1 +155 -32

toasterson commented 2026-10-16 09:49:06 +00:00

Owner

Copy link

Fixes #662.

The unknown-job sweep test could age its supposedly fresh fixture past the 1,000 ms threshold while redb committed or the runner was descheduled. This change passes one clock sample through a private resolve_stale_unknown_jobs_at(now) implementation and uses it to initialize missing unknown_since records too. Production continues to call the public wrapper with the real clock; expiry thresholds, ownership fencing, and reconciliation policy are unchanged.

Tests use fixed logical timestamps for threshold-minus-one, exact-threshold, past-threshold, and future clocks. They check missing-clock initialization without reset, repeat sweeps, archived receipt visibility and side-key cleanup, restart preservation, active workers, foreign runtimes, Internal jobs, and disposable jobs (including removed profiles with retained generation receipts). The fixture threshold remains 1,000 ms. There are no added sleeps or wider timeouts.

Validation:

• cargo test -p akh-session-services --no-default-features --features runtime --lib: 79 passed.
• Sweep tests repeated in 20 trials across four concurrent processes: 100/100 test executions passed.
• File-scoped rustfmt and git diff --check: passed.
• CI full session-platform feature-set library tests: 173 passed, 0 failed, 6 ignored (original Solstice task 37356, request 32070818-2d96-4a55-b6e7-d86bc3a0eba9). Whole-run CI: 15/15 jobs passed in run 4306 at cc705e4b6c5201b7b2d18d36e1318858d6d111fa, including the complete platform storage harness and all dependent image checks. No retry or timeout changes were needed.

This is the separately authorized follow-up after Talos #220 / platform-storage PR 6 and before PR 7. It does not change manifests or activate workloads. No local Docker harness was run because this change touches only the runtime clock and its tests; local builds use the dedicated lean target, without candle/core feature compilation.

Issue #662 records three distinct incidents: the sweep timing failure, an analogy failure, and the SQLite contention fixture subsequently fixed in #661. This PR fixes the sweep only; it does not claim to fix the analogy test.

Fixes #662. The unknown-job sweep test could age its supposedly fresh fixture past the 1,000 ms threshold while redb committed or the runner was descheduled. This change passes one clock sample through a private `resolve_stale_unknown_jobs_at(now)` implementation and uses it to initialize missing `unknown_since` records too. Production continues to call the public wrapper with the real clock; expiry thresholds, ownership fencing, and reconciliation policy are unchanged. Tests use fixed logical timestamps for threshold-minus-one, exact-threshold, past-threshold, and future clocks. They check missing-clock initialization without reset, repeat sweeps, archived receipt visibility and side-key cleanup, restart preservation, active workers, foreign runtimes, Internal jobs, and disposable jobs (including removed profiles with retained generation receipts). The fixture threshold remains 1,000 ms. There are no added sleeps or wider timeouts. Validation: - `cargo test -p akh-session-services --no-default-features --features runtime --lib`: 79 passed. - Sweep tests repeated in 20 trials across four concurrent processes: 100/100 test executions passed. - File-scoped rustfmt and `git diff --check`: passed. - CI full session-platform feature-set library tests: 173 passed, 0 failed, 6 ignored (original Solstice task 37356, request `32070818-2d96-4a55-b6e7-d86bc3a0eba9`). Whole-run CI: **15/15 jobs passed** in [run 4306](https://code.aopc.cloud/toasterson/akh-medu/actions/runs/4306) at `cc705e4b6c5201b7b2d18d36e1318858d6d111fa`, including the complete platform storage harness and all dependent image checks. No retry or timeout changes were needed. This is the separately authorized follow-up after Talos #220 / platform-storage PR 6 and before PR 7. It does not change manifests or activate workloads. No local Docker harness was run because this change touches only the runtime clock and its tests; local builds use the dedicated lean target, without candle/core feature compilation. Issue #662 records three distinct incidents: the sweep timing failure, an analogy failure, and the SQLite contention fixture subsequently fixed in #661. This PR fixes the sweep only; it does not claim to fix the analogy test.

toasterson added 1 commit 2026-10-16 09:49:08 +00:00

fix(runtime): use deterministic time in unknown-job sweep tests cc705e4b6c

toasterson merged commit 96dd04d9dc into main 2026-10-16 10:24:30 +00:00

toasterson referenced this pull request from a commit 2026-10-16 10:24:32 +00:00

fix(runtime): make unknown-job sweep tests deterministic (#663)

toasterson deleted branch fix/runtime-sweep-deterministic-clock 2026-10-16 10:24:37 +00:00

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

follow-up

Documented operational follow-up

No labels

follow-up

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

toasterson/akh-medu!663

No description provided.