Answer every seat call within ten seconds and keep what came of it (hq issue 265) #72

2026-10-05T23:15:15Z

mesh-admin commented

2026-10-05 23:15:15 +00:00

novox/hq issue 265.

Cause. A seat verb answered only when its command finished. A push takes about 27 s (median), and the console waits 30 s, so slower pushes were reported as "did not answer in time" even though they had pushed. Worse, a push sends the bus holder its changed user list first. The broker reloads, which re-registers every user and empties the server's table of replies it may send, so the push's eventual answer is refused (Permissions Violation for Publish to "_INBOX.…"). Every one of the 30 journal lines came 35 to 50 s after a timed-out push, right after a broker reload.

Change.

internal/link/calls.go: each call is answered once, in full when it finishes within AnswerWithin (10 s), otherwise as {running, call: <id>}. The command runs on and its answer is kept (last 100 calls, settings and secrets never kept, large in-time answers not kept). Acknowledge(ctx) answers at once. The connection's error handler is chained, so a refused answer is recorded on its call and logged with the call's id.
seatverbs.go: push (and command push …) acknowledges before running. The new in-process verb calls lists the recent calls, or returns one whole by id.
catalogue/verbs.go: adds calls, and push's description says it answers at once with an id.
broker/nats.go: ResponseTTL is named, and the rendered config is unchanged (golden passes).

Tests. internal/link/calls_test.go covers answering in time, answering as running with the answer kept, acknowledgement, refusal recorded, no secrets kept, and the ring. cmd/mesh-controller/calls_verb_test.go covers push answering first, calls being served and its refusals, and AnswerWithin staying under both the console wait and ResponseTTL. gofmt, go vet ./..., go test ./... are clean (race -count=3 on the link tests).

Merge with or before mesh-tools fix/a-timeout-is-not-a-failure. The order between the two does not matter for correctness.

novox/hq issue 265. **Cause.** A seat verb answered only when its command finished. A push takes about 27 s (median), and the console waits 30 s, so slower pushes were reported as "did not answer in time" even though they had pushed. Worse, a push sends the bus holder its changed user list first. The broker reloads, which re-registers every user and empties the server's table of replies it may send, so the push's eventual answer is refused (`Permissions Violation for Publish to "_INBOX.…"`). Every one of the 30 journal lines came 35 to 50 s after a timed-out push, right after a broker reload. **Change.** - `internal/link/calls.go`: each call is answered once, in full when it finishes within `AnswerWithin` (10 s), otherwise as `{running, call: <id>}`. The command runs on and its answer is kept (last 100 calls, settings and secrets never kept, large in-time answers not kept). `Acknowledge(ctx)` answers at once. The connection's error handler is chained, so a refused answer is recorded on its call and logged with the call's id. - `seatverbs.go`: `push` (and `command push …`) acknowledges before running. The new in-process verb `calls` lists the recent calls, or returns one whole by id. - `catalogue/verbs.go`: adds `calls`, and push's description says it answers at once with an id. - `broker/nats.go`: `ResponseTTL` is named, and the rendered config is unchanged (golden passes). **Tests.** `internal/link/calls_test.go` covers answering in time, answering as running with the answer kept, acknowledgement, refusal recorded, no secrets kept, and the ring. `cmd/mesh-controller/calls_verb_test.go` covers push answering first, `calls` being served and its refusals, and AnswerWithin staying under both the console wait and ResponseTTL. gofmt, go vet ./..., go test ./... are clean (race -count=3 on the link tests). Merge with or before mesh-tools `fix/a-timeout-is-not-a-failure`. The order between the two does not matter for correctness.

mesh-admin added 1 commit 2026-10-05 23:15:15 +00:00

Answer every seat call within ten seconds and keep what came of it (hq issue 265) 801552c0eb

A push outlasted the console's 30s wait and, when it sent the bus its
changed user list, the broker's reload forgot the reply it may send:
the push happened and its caller was told it did not answer. Calls now
answer in full or as running with an id, a push answers before it
sends, refused answers are recorded on their call, and 'calls' reads
them back.

mesh-admin merged commit eda457f415 into main

2026-10-05 23:33:11 +00:00

mesh-admin deleted branch fix/a-verb-answers-before-its-caller-gives-up

2026-10-05 23:33:11 +00:00

mesh-admin commented

2026-10-06 23:10:28 +00:00

Delivery novox/mesh-controller@eda457f41566 — held since 2026-10-06T23:05:28Z

Held for a person: merged without a passing check: only a person decides that it goes on — mesh-delivery.release with why lets it go on.

Delivery plan — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step

tier 0: mesh-controller
tier 1: build-agent
tier 2: route-proxy
ace: receives build-agent, route-proxy
g14: receives build-agent
novox: receives mesh-controller, build-agent, route-proxy
shanks: receives build-agent
build-agent keeps data (ADR 0233): what it holds is backed up before it moves
route-proxy provides route: its consumers are sent again after it
route-proxy keeps data (ADR 0233): what it holds is backed up before it moves

Transitions

2026-10-06 23:05 (new) → proposed (announced): novox/mesh-controller#72 merged; its head was not heard by this owner, so it is made from the forge's word
2026-10-06 23:05 proposed → held (merged-unchecked): merged without a passing check — a verdict known with it is taken first: only a person decides that it goes on

The commit's note under refs/notes/mesh-plan keeps every transition: git log --notes=mesh-plan.

**Delivery** `novox/mesh-controller@eda457f41566` — **held** since 2026-10-06T23:05:28Z Held for a person: merged without a passing check: only a person decides that it goes on — `mesh-delivery.release` with why lets it go on. **Delivery plan** — builds build-agent, mesh-controller, route-proxy → ace, g14, novox, shanks; no bus step - tier 0: mesh-controller - tier 1: build-agent - tier 2: route-proxy - ace: receives build-agent, route-proxy - g14: receives build-agent - novox: receives mesh-controller, build-agent, route-proxy - shanks: receives build-agent - build-agent keeps data (ADR 0233): what it holds is backed up before it moves - route-proxy provides route: its consumers are sent again after it - route-proxy keeps data (ADR 0233): what it holds is backed up before it moves **Transitions** - 2026-10-06 23:05 (new) → proposed (announced): novox/mesh-controller#72 merged; its head was not heard by this owner, so it is made from the forge's word - 2026-10-06 23:05 proposed → held (merged-unchecked): merged without a passing check — a verdict known with it is taken first: only a person decides that it goes on The commit's note under `refs/notes/mesh-plan` keeps every transition: `git log --notes=mesh-plan`.

This repo is archived. You cannot comment on pull requests.

1 Participants

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: novox/mesh-controller#72