[BUG] Named idle time-resolution VM exceeds its original QEMU deadline #75

Open
opened 2026-09-23 22:35:30 +02:00 by erikinkinen · 0 comments
Owner

Summary

The maintained named idle time-resolution scenario exceeds its original 60-second QEMU deadline in an existing CI push run. The full catalog finishes with 488 of 489 scenarios passing. The cause is not established; this is distinct from the ext4 deadlines in the paired PR run.

Reproduction

Original push run 1751, Integration 7aa451477c8ac5238bb6c334b69a466b2a94996b, runs the existing catalog with twelve VM workers. The failed contract can be selected with:

bash scripts/run-scenario.sh tests/scenarios/subsystem-named-idle-time-resolution-positive.toml components.toml

One original CI failure is retained. Standalone reproducibility and a last known good revision are not established. The failed run was not retried.

Expected and actual behavior

Expected: target-scoped timed resolution wakes the idle multi-endpoint named server, the client exits successfully with ERIX_KERNEL:PROC_EVENT_STATUS=0000000000000000, and the scenario completes within its unchanged 60-second deadline and exit contract.

Actual: scenario status 1, error: qemu timed out after 60s. The captured serial output lacks the required zero process-exit status and contains none of this scenario's forbidden fault markers. Its last complete EriX line is ERIX_ROOTD:READY:STORAGE, followed by a partial logging marker and the runner's stderr separator. These lines do not establish the executing instruction or root cause. QEMU receives signal 15 from the timeout runner.

Environment and identity

  • Exact Integration graph and workflow are linked above; API run 10369, integration job 22260, VM worker 6.
  • Release subsystem-positive image, x86-64 QEMU/OVMF, one virtual CPU, original 60-second scenario deadline.
  • Maintained Rootd time probe, Named and Timed smoke services, original scenario capability grants and runtime configuration.
  • The original scenario policy has SHA-256 be3fe4d080f0583e3fae07a2483650f809587b5bd18a62e9a324d16f9545b559.
  • Separate packaged image digest and last known working revision are not established by the retained CI logs.

Evidence and prior checks

Original CI review — 29 September 2026: run 1826 passes the complete workflow, all 492 VM scenarios, canonical dev/release image checks and subsequent physical interactive gates. The paired run 1825 passes 491/492 scenarios and fails only subsystem-named-idle-time-resolution-positive at its original 60-second QEMU deadline. This is another occurrence of the existing Named timeout report, whose cause remains unresolved. Different outcomes on the same revision do not establish a fix.

Runs 1827 and 1828, on the intermediate receipt-owner revision, fail in native_memory_fixture.py before VM execution: the selected memory producer rejects the changed process-owner identity. This is the known coordinated-adoption dependency; signed Integration e4ffe2efcf4f06386ffc72d227c4d9c9646e86d1 adopts the matching Lib-cstd producer. Its original CI 1829/1830 remains running; the later roadmap-only CI 1831/1832 remains waiting. The intermediate refusal is retained rather than bypassed or retried.

All twelve newly terminal logs are complete and hash-checked, totaling 27,929,343 bytes with zero warning candidates. Rust and Markdown jobs pass in all four runs. The separate Docs failed logs remain unavailable with HTTP 500. No unchanged failure was rerun and no additional lifecycle acceptance is claimed.

Three complete original terminal job logs are retained with verified lengths and hashes: 13,834,471 bytes and zero warning candidates. The integration log SHA-256 is ae0a752fe1468ba1b5aeded03a96539d5e2a83f3647f9f8972493d9b7df40399. Rust 320/321-test configurations and Markdown pass. Both allocator scenarios pass; later canonical image and physical-input gates are not reached after catalog failure.

The paired PR run 1752 fails different ext4 scenarios, recorded in the existing report. Different outcomes at the same revision do not identify or repair either cause. Investigation must retain the original timing, exit, marker and authority requirements. No deadline widening or unchanged-input retry is accepted as a correction.

Tracking

Track investigation in Integration PR 12, integration validation and Phase 6 completion. This report remains open pending an attributed correction and unchanged-contract regression evidence.

## Summary The maintained named idle time-resolution scenario exceeds its original 60-second QEMU deadline in an existing CI push run. The full catalog finishes with 488 of 489 scenarios passing. The cause is not established; this is distinct from the ext4 deadlines in the paired PR run. ## Reproduction Original [push run 1751](https://git.erikinkinen.fi/erix/integration/actions/runs/1751), Integration `7aa451477c8ac5238bb6c334b69a466b2a94996b`, runs the existing catalog with twelve VM workers. The failed contract can be selected with: ```sh bash scripts/run-scenario.sh tests/scenarios/subsystem-named-idle-time-resolution-positive.toml components.toml ``` One original CI failure is retained. Standalone reproducibility and a last known good revision are not established. The failed run was not retried. ## Expected and actual behavior Expected: target-scoped timed resolution wakes the idle multi-endpoint named server, the client exits successfully with `ERIX_KERNEL:PROC_EVENT_STATUS=0000000000000000`, and the scenario completes within its unchanged 60-second deadline and exit contract. Actual: scenario status 1, `error: qemu timed out after 60s`. The captured serial output lacks the required zero process-exit status and contains none of this scenario's forbidden fault markers. Its last complete EriX line is `ERIX_ROOTD:READY:STORAGE`, followed by a partial logging marker and the runner's stderr separator. These lines do not establish the executing instruction or root cause. QEMU receives signal 15 from the timeout runner. ## Environment and identity - Exact Integration graph and workflow are linked above; API run 10369, integration job 22260, VM worker 6. - Release subsystem-positive image, x86-64 QEMU/OVMF, one virtual CPU, original 60-second scenario deadline. - Maintained Rootd time probe, Named and Timed smoke services, original scenario capability grants and runtime configuration. - The original scenario policy has SHA-256 `be3fe4d080f0583e3fae07a2483650f809587b5bd18a62e9a324d16f9545b559`. - Separate packaged image digest and last known working revision are not established by the retained CI logs. ## Evidence and prior checks Original CI review — 29 September 2026: [run 1826](https://git.erikinkinen.fi/erix/integration/actions/runs/1826) passes the complete workflow, all 492 VM scenarios, canonical dev/release image checks and subsequent physical interactive gates. The paired [run 1825](https://git.erikinkinen.fi/erix/integration/actions/runs/1825) passes 491/492 scenarios and fails only `subsystem-named-idle-time-resolution-positive` at its original 60-second QEMU deadline. This is another occurrence of [the existing Named timeout report](https://git.erikinkinen.fi/erix/integration/issues/75), whose cause remains unresolved. Different outcomes on the same revision do not establish a fix. [Runs 1827](https://git.erikinkinen.fi/erix/integration/actions/runs/1827) and [1828](https://git.erikinkinen.fi/erix/integration/actions/runs/1828), on the intermediate receipt-owner revision, fail in `native_memory_fixture.py` before VM execution: the selected memory producer rejects the changed process-owner identity. This is the known coordinated-adoption dependency; signed Integration `e4ffe2efcf4f06386ffc72d227c4d9c9646e86d1` adopts the matching Lib-cstd producer. Its original CI 1829/1830 remains running; the later roadmap-only CI 1831/1832 remains waiting. The intermediate refusal is retained rather than bypassed or retried. All twelve newly terminal logs are complete and hash-checked, totaling 27,929,343 bytes with zero warning candidates. Rust and Markdown jobs pass in all four runs. The separate Docs failed logs remain unavailable with HTTP 500. No unchanged failure was rerun and no additional lifecycle acceptance is claimed. Three complete original terminal job logs are retained with verified lengths and hashes: 13,834,471 bytes and zero warning candidates. The integration log SHA-256 is `ae0a752fe1468ba1b5aeded03a96539d5e2a83f3647f9f8972493d9b7df40399`. Rust 320/321-test configurations and Markdown pass. Both allocator scenarios pass; later canonical image and physical-input gates are not reached after catalog failure. The paired [PR run 1752](https://git.erikinkinen.fi/erix/integration/actions/runs/1752) fails different ext4 scenarios, recorded in [the existing report](https://git.erikinkinen.fi/erix/integration/issues/73). Different outcomes at the same revision do not identify or repair either cause. Investigation must retain the original timing, exit, marker and authority requirements. No deadline widening or unchanged-input retry is accepted as a correction. ## Tracking Track investigation in [Integration PR 12](https://git.erikinkinen.fi/erix/integration/pulls/12), [integration validation](https://git.erikinkinen.fi/erix/integration/issues/3) and [Phase 6 completion](https://git.erikinkinen.fi/erix/integration/issues/65). This report remains open pending an attributed correction and unchanged-contract regression evidence.
erikinkinen added this to the Alpha 1 milestone 2026-10-30 09:39:22 +01:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
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
erix/integration#75
No description provided.