dohertj2/lmxopcua

Out-of-process script soft deadline includes pipe round trip and worker scheduling delay #827

New Issue

Closed

opened 2026-11-11 00:05:59 -05:00 by dohertj2 · 1 comment

dohertj2 commented

2026-11-11 00:05:59 -05:00

Owner

Found during #811. The worker's soft deadline (= seam run budget) starts at submit, so a starved worker process under load reports "script timed out" for trivial scripts; in Shadow mode that appears as a script mismatch and skews the census used to decide the OutOfProcess flip (#780). The in-process guard excludes thread-pool starvation. Decide whether to start the clock when the worker begins the evaluation (and report queue/transport delay separately). Also audit the other ScriptWorker tests for tight deadlines under load.

Found during #811. The worker's soft deadline (= seam run budget) starts at submit, so a starved worker process under load reports "script timed out" for trivial scripts; in Shadow mode that appears as a script mismatch and skews the census used to decide the OutOfProcess flip (#780). The in-process guard excludes thread-pool starvation. Decide whether to start the clock when the worker begins the evaluation (and report queue/transport delay separately). Also audit the other ScriptWorker tests for tight deadlines under load.

dohertj2 added the type:followup priority:p3 labels 2026-11-11 00:05:59 -05:00

dohertj2 referenced this issue

2026-11-11 00:49:03 -05:00

ScriptWorker ShadowModeTests: one unexplained wrong sibling value under heavy CPU load #811

dohertj2 referenced this issue

2026-11-11 03:27:28 -05:00

Script worker: Windows suspend path (NtSuspendProcess) for the soft-clock tests is untested #842

dohertj2 referenced this issue from a commit

2026-11-11 05:24:23 -05:00

fix(scripting): out-of-process soft deadline measures the script's own run, not the pipe or a starved worker (#827)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:24 -05:00

test(scriptworker): tight-heartbeat tests warm the worker before arming the 3-interval silence kill (#827 audit)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:24 -05:00

fix(scripting): soft notice resumes the caller on the dispatcher; a late notice still cancels and counts (#827 review F1, F2)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:24 -05:00

fix(scripting): host progress ceiling kills a worker that heartbeats but never answers (#827 review F3)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:24 -05:00

fix(scripting): count re-triggers answered 'worker unresponsive' (#827 review F5)

dohertj2 referenced this issue from a commit

2026-11-11 05:24:24 -05:00

docs(scripting): T starts at the run's start but is wall clock from there; document the late-notice, ceiling and counter rules (#827 review F4)

dohertj2 commented

2026-11-11 05:24:33 -05:00

Author

Owner

Fixed (merge of fix/827-worker-deadline). Pushed to master in d1b6cad2. The soft deadline is the worker's clock from run start (negotiated in Hello/Ready); late soft notices cancel and count the timeout; host transport deadline T+G and progress ceiling killed_unresponsive = 3×max(T+G) + silence limit; timeouts metric tagged answer=stalled|retrigger. Windows suspend path untested: #842.

Fixed (merge of fix/827-worker-deadline). Pushed to master in `d1b6cad2`. The soft deadline is the worker's clock from run start (negotiated in Hello/Ready); late soft notices cancel and count the timeout; host transport deadline T+G and progress ceiling `killed_unresponsive` = 3×max(T+G) + silence limit; timeouts metric tagged `answer=stalled|retrigger`. Windows suspend path untested: #842.

dohertj2 closed this issue

2026-11-11 05:24:33 -05:00

dohertj2 referenced this issue

2026-11-11 05:24:34 -05:00

Runtime.Tests: DriverHostActorSourceRemovedTests late-burst test flaked once #828

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: dohertj2/lmxopcua#827