feat(events): the runner (Phase 2)
All checks were successful
PR Checks / bot-tests (pull_request) Successful in 32s
PR Checks / client-build (pull_request) Successful in 33s
PR Checks / server-tests (pull_request) Successful in 5m29s

`utils/eventRunner.js`, the eighth poller, wired into server.js beside
engagementWorker. Its tick reclaims stale leases, sweeps occurrences past their
grace window into `missed`, advances each due run through its phases, and drains
that phase's steps in `seq` order. The three core actions from Phase 1 get real
bodies, so a published event started from the existing run route now announces,
waits and completes on its own.

No routes are added: a runner has no surface, and the live controls stay Phase
3's.

Four things the org lead settled (2026-09-02): a parked step is `running` with a
NULL lease; `await: 'human'` and `holdFor` are ordinary success-envelope members
rather than special cases keyed on an action id; a run whose concurrency key is
held stays `scheduled` and lets its grace window decide; and `n` in §L's
`retry(n)` is a runner constant.

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
2026-09-02 06:32:24 -05:00
parent d88906e43c
commit 2e964cfeee
10 changed files with 2527 additions and 41 deletions

View File

@@ -13,28 +13,20 @@
// a human to go and do something. A deployment with no game module installed has
// a working event system made of exactly these.
//
// **Nothing here dispatches yet.** Phase 1 builds the registry, the id grammar,
// the risk classes and the param validation; Phase 2 builds `utils/eventRunner.js`
// and is what calls `perform()`. The bodies below therefore answer with the
// envelope §F defines for a refusal — and specifically NOT with `{ ok: true }`,
// which is the one wrong answer a placeholder can give: `ok: true` on an action
// that did nothing is a recorded world change that did not occur, which is the
// exact mistake the envelope's failure default exists to prevent. `retry: false`
// because a missing runner is not a transient condition.
// **Phase 2 gave all three real bodies**, and between them they exercise every
// shape §F's envelope can take: `core.announce` does work and finishes,
// `core.wait` finishes while deferring what follows it, and `core.cue` succeeds
// without finishing at all. The runner learns nothing about any of them by id —
// each says what it needs in the envelope, through the same two members Phase 7
// hands to a module.
//
// **This file must not touch the database.** It is required from `registerCore()`,
// which runs under `routeManifest.js` and `swagger.js` against a dead pool
// (MODULE_API.md §2.2). It is pure data plus three functions that are not called.
// (MODULE_API.md §2.2). Nothing below runs at require time; the announce leg is
// looked up inside `perform()`, per call, which is also what makes a leg
// registered by a module that booted later reachable at all.
// A factory rather than one shared function, because `perform`'s argument is
// §F's dispatch envelope — `{ runId, stepId, idempotencyKey, scope, params,
// actor, verify }` — and it does not carry the action's own id. Closing over it
// is what lets the refusal name which action refused.
const notWiredYet = (actionId) => async () => ({
ok: false,
retry: false,
error: `${actionId} is declared in Phase 1 and dispatched from Phase 2`,
})
const registries = require('../modules/registries')
const ACTIONS = [
{
@@ -81,7 +73,52 @@ const ACTIONS = [
},
],
perform: notWiredYet('core.announce'),
/**
* Publish through the announce leg the step names.
*
* **The legs are reused rather than reimplemented** (§J, "reuse the legs"):
* `discord` is core's and `towncrier` is module-uo's, both already registered,
* both already carrying a `classify()` that knows what their transport's
* failures mean. An event announcement that went out by some other path would
* be a second delivery mechanism with its own bugs.
*
* A leg's `dispatch()` takes a POST — that is the shape the news path gave it
* — so an event announcement is presented as one. `excerpt` is the body
* because it is the field every leg renders as prose, and `image_url` is null
* because an event announcement has no article behind it to illustrate.
* Widening the leg contract to carry a second payload shape is a
* MODULE_API change, and Phase 7 is where those are made.
*
* The leg id is checked HERE rather than at authoring time, and that is not
* laxness: legs are registered by modules, and a spec is validated in a
* process that may have booted before the module that owns the leg.
*/
async perform({ params, verify }) {
const registered = registries.announceLeg(params.leg)
if (!registered) {
// Terminal, not transient. A leg nobody registers will not appear
// between two attempts sixty seconds apart, and the honest cause — a
// module removed, or a typo the authoring form could not catch — is a
// thing a human fixes.
return { ok: false, retry: false, error: `no module registers the announce leg "${params.leg}"` }
}
// A dry run reports what it WOULD do and sends nothing (§I). Answering
// before the dispatch rather than inside the leg is what keeps that true
// for legs written by people who never read this file.
if (verify) return { ok: true }
const result = await registered.dispatch({
title: params.title || null,
excerpt: params.body,
image_url: null,
})
// The leg's own classification, not a second opinion. `retry` vs
// `terminal` for a Discord webhook is a judgement `discordAnnounce.classify`
// already makes, and making it twice is how the two drift.
const { outcome, error } = registered.classify(result)
if (outcome === 'done') return { ok: true }
return { ok: false, retry: outcome === 'retry', error: error || `announce leg "${params.leg}" refused` }
},
},
{
@@ -105,11 +142,18 @@ const ACTIONS = [
},
],
// A wait is a genuine no-op at dispatch, and it will stay one: the delay is
// the NEXT step's `due_at`, which the runner owns, not something this
// function sleeps through. A `perform` that slept would hold a step's claim
// for the duration and turn a five-minute pause into a five-minute lease.
perform: notWiredYet('core.wait'),
// A wait is a genuine no-op at dispatch, and it stayed one: the delay is the
// NEXT step's `due_at`, which the runner owns, not something this function
// sleeps through. A `perform` that slept would hold a step's claim for the
// duration and turn a five-minute pause into a five-minute lease — and the
// reclaim would then re-dispatch it, so a long enough wait would never end.
//
// `holdFor` is an ordinary envelope member (org lead, 2026-09-02), which is
// why the runner can honour this without knowing what `core.wait` is.
async perform({ params, verify }) {
if (verify) return { ok: true }
return { ok: true, holdFor: params.seconds }
},
},
{
@@ -142,11 +186,26 @@ const ACTIONS = [
},
],
// Phase 2 gives this its parking semantics — a cue step does not complete
// when `perform` answers, it completes when a human presses confirm, and the
// control that does so is Phase 3's. Both of those are what make this the
// one action whose runtime shape is deliberately not decided here.
perform: notWiredYet('core.cue'),
/**
* Post the instruction and PARK. The step does not complete here.
*
* `await: 'human'` is the envelope member that says so (org lead,
* 2026-09-02), and the runner's answer to it is to leave the step `running`
* with a NULL lease — genuinely in flight, nothing holding it, so the stale
* reclaim passes it by and a cue posted on Friday is still waiting on Monday.
* The step ends when someone presses confirm, which is Phase 3's control.
*
* **Nothing is delivered from here in Phase 2, and that is visible rather
* than pretended.** The instruction is carried by the step's own params and
* shown on the run console; routing it to Discord or to a staff inbox is
* Phase 10's integration work, through the engagement triggers that own every
* other notification on this platform. An action that grew its own delivery
* path would be the second one.
*/
async perform({ verify }) {
if (verify) return { ok: true }
return { ok: true, await: 'human' }
},
},
]

View File

@@ -0,0 +1,166 @@
// ── Dispatching one step to one action ─────────────────────────────────────
//
// EVENTS.md §F. This is the boundary between the runner and code core did not
// write, and it exists as its own file because it has exactly one job: call
// `perform()` and turn whatever comes back — an envelope, a lie, a throw, a
// promise that never settles — into one of four classifications the runner knows
// how to act on.
//
// **§F's load-bearing rule, and the reason none of this is inlined into the
// runner: no shape a failure can take may read as success.** A rejected promise,
// a throw, a timeout, a non-object and a missing `ok` are all
// `{ ok: false, retry: true }`. That is the inverse of `registerTeamProvider`'s
// default, deliberately — a team provider that refuses leaves core showing what
// it already had, because staleness is cheap, whereas an action that half-ran and
// was recorded as `done` is a world change nothing will ever come back for.
//
// **The timeout is the module contract's, not this file's opinion.** Every action
// declares `budgetMs` at registration and the registry bounds it there; here it
// is enforced. Without it a module whose `perform()` awaits a socket that never
// answers holds a step's claim until the lease expires, and the reclaim then
// re-dispatches it — which is how one wedged sidecar becomes an infinite loop
// rather than a failed step.
const registries = require('../modules/registries')
const log = require('../utils/logger')('events')
// What a classification can be. `parked` is Phase 2's addition and it is the one
// outcome that is neither terminal nor a retry: the action succeeded, and the
// step is not finished, because something outside this system has to happen next.
const OUTCOMES = ['done', 'parked', 'retry', 'terminal']
// The upper bound on `holdFor`, in seconds. A wait is a scheduling instruction,
// not a lease, so this is generous — but it is bounded, because an action that
// answers `holdFor: 1e9` would park the phase past the heat death of the shard
// and the step that did it would look, in the console, exactly like one that
// worked.
const MAX_HOLD_SECONDS = 7 * 24 * 60 * 60
/**
* Run `fn()` under a deadline.
*
* The loser of the race is not cancelled — JavaScript has no such thing, and a
* `perform()` still awaiting a socket keeps awaiting it. What the deadline buys
* is that the RUNNER stops waiting, which is the half that matters: the step is
* classified, the claim is released, and the tick moves on. A late answer from
* the abandoned call lands on a step that has already been written, and the
* idempotency key is what makes the retry that follows safe on the game side.
*/
function withDeadline(fn, ms, actionId) {
let timer = null
const deadline = new Promise((resolve) => {
timer = setTimeout(
() => resolve({ __timedOut: true, error: `${actionId} exceeded its ${ms}ms budget` }),
ms,
)
if (timer.unref) timer.unref()
})
return Promise.race([Promise.resolve().then(fn), deadline]).finally(() => {
if (timer) clearTimeout(timer)
})
}
/**
* Turn a raw `perform()` answer into `{ outcome, error?, holdSeconds?, resources? }`.
*
* Exported and pure, so the classification rules are testable without a registry,
* a database or a clock — which matters because they are the rules that decide
* whether a world change is recorded as having happened.
*/
function classify(result, actionId) {
if (result && result.__timedOut) {
// Transient by default: a timeout says nothing about whether the action ran.
// That ambiguity is exactly what the idempotency key exists to resolve, and
// resolving it on the game side is Phase 11's protocol work — until then a
// retry is the honest choice and the risk class decides what happens when the
// retries run out.
return { outcome: 'retry', error: result.error }
}
if (result === null || typeof result !== 'object' || Array.isArray(result)) {
return { outcome: 'retry', error: `${actionId} answered with no envelope` }
}
if (result.ok !== true) {
// `retry` must be opted into. An action that means "this will never work"
// says `retry: false`, and an envelope that forgot to say anything gets the
// benefit of the doubt on the transient question but not on the success one.
const retry = result.retry !== false
return {
outcome: retry ? 'retry' : 'terminal',
error: result.error ? String(result.error) : `${actionId} refused`,
}
}
// ── The two success shapes that are not "finished" ──
//
// Both were settled by the org lead on 2026-09-02, and both are envelope
// members rather than special cases keyed on an action id, so that the runner
// never names a verb. `core.cue` and `core.wait` reach them through the same
// door Phase 7 opens to a module's own long-running action.
if (result.await === 'human') {
return { outcome: 'parked', error: null, resources: result.resources || [] }
}
let holdSeconds = 0
if (result.holdFor !== undefined && result.holdFor !== null) {
const n = Number(result.holdFor)
if (!Number.isFinite(n) || n < 0) {
return { outcome: 'terminal', error: `${actionId} answered a bad holdFor "${result.holdFor}"` }
}
holdSeconds = Math.min(Math.floor(n), MAX_HOLD_SECONDS)
}
return { outcome: 'done', error: null, holdSeconds, resources: result.resources || [] }
}
/**
* Dispatch one step. Never throws.
*
* `verify` rides through to `perform()` unchanged (§I's dry run, Phase 6's
* route): `verify === true` means validate and report, change nothing. It is
* passed from here rather than being a separate code path so that the dry run
* exercises the real dispatcher — a dry run down a second path is a dry run of
* the second path.
*/
async function dispatchStep(step, { run, actor = null, verify = false } = {}) {
const action = registries.eventAction(step.action_id)
if (!action) {
// §L, verbatim: "a step naming one fails terminal with the module named, and
// the run degrades rather than claiming success. Never a silent skip." The
// module was uninstalled or failed to boot between publish and now — publish
// refuses a dormant step, so this cannot be an authoring mistake.
return { outcome: 'terminal', error: `no module registers "${step.action_id}"`, dormant: true }
}
const envelope = {
runId: run.id,
stepId: step.id,
idempotencyKey: step.idempotency_key,
scope: run.scope || '',
params: step.params || {},
actor,
verify: Boolean(verify),
}
let raw
try {
raw = await withDeadline(() => action.perform(envelope), action.budgetMs, action.id)
} catch (err) {
// A module should not throw, and if one does it is a transient failure rather
// than a crashed tick — announceWorker's posture with its legs, and the
// reason one bad module cannot stop every other run on the deployment.
log.warn('event action threw', { action: action.id, run: run.id, step: step.id, message: err.message })
return { outcome: 'retry', error: err.message }
}
const classification = classify(raw, action.id)
if (step.action_version && action.version !== step.action_version) {
// Not a refusal: the step was authored against an older declaration and the
// module has moved on. The editor is where that becomes a warning (§F); here
// it is recorded, so a run that behaved oddly can be explained afterwards by
// reading the log rather than by guessing.
classification.actionVersionDrift = { authored: step.action_version, registered: action.version }
}
return classification
}
module.exports = { dispatchStep, classify, withDeadline, OUTCOMES, MAX_HOLD_SECONDS }

View File

@@ -23,6 +23,16 @@ const KINDS = [
'phase.entered', // a phase's steps were materialised
'step.status', // a step transition, with the module's answer
'note', // a human action taken from the admin surface
// Phase 2's, all five of them answers to a question an operator asks out
// loud. `run.blocked` in particular is the whole reason this table exists
// rather than a server log line: "it did not start because run 37 holds
// invasion:Yew" is a fact with two run ids in it, and it has to be
// queryable from the run that did NOT start.
'run.blocked', // an occurrence held off: another run has its concurrency key
'run.health', // a health change, which is not a status change
'step.retry', // a step failed transiently and will be attempted again
'step.parked', // a step is waiting on a human and nothing is holding it
'phase.completed', // every step of a phase reached a terminal status
]
const hydrate = (row) => row && { ...row, detail: parseJson(row.detail, null) }
@@ -64,4 +74,33 @@ async function write({ runId, stepId = null, kind, phase = null, detail = null }
}
}
module.exports = { KINDS, listForRun, write }
/**
* Delete log lines belonging to runs that are both TERMINAL and older than
* `before`, a bounded number at a time.
*
* The schema comment beside `idx_evlog_at` parked this sweep here, and it is the
* rule Engagement Phase 14 arrived at applied to a second high-cardinality table:
* **only terminal rows are eligible.** A run still in flight keeps every line it
* has, however old — the log's whole job is answering "why didn't phase 3 start?"
* about a run that is, right now, not starting phase 3, and a horizon that could
* reach a live run would delete the answer while the question was still open.
*
* `LIMIT` makes one call a bounded amount of work rather than a table-sized
* transaction; the timer runs again and takes the next slice. The join is on the
* run's terminal status rather than on a precomputed id list so that a run which
* reached a terminal state between the two would not be missed.
*/
const pruneTerminal = async (before, limit = 5000) => {
const n = Math.min(Math.max(Number(limit) || 5000, 1), 50_000)
const result = await query(
`DELETE l FROM event_run_log l
JOIN event_runs r ON r.id = l.run_id
WHERE r.status IN ('completed','cancelled','failed','missed')
AND COALESCE(r.ended_at, r.updated_at) < ?
LIMIT ${n}`,
[before],
)
return Number(result?.affectedRows || 0)
}
module.exports = { KINDS, listForRun, write, pruneTerminal }

View File

@@ -92,4 +92,201 @@ const statusCounts = async (runId) => {
return Object.fromEntries(rows.map((r) => [r.status, Number(r.n)]))
}
module.exports = { listForRun, getById, materialisePhase, statusCounts, idempotencyKey }
// ── Phase 2: draining a step ───────────────────────────────────────────────
//
// The CAS claim, the lease, the attempt counter and the terminal writes. Phase 1
// left all of it out rather than stubbing it, and this is where it lands.
//
// **Two rules govern everything below, and both were paid for once already.**
//
// 1. `attempts` is incremented by the CLAIM and by nothing else, and no recovery
// path ever resets it. Engagement Phase 14's defect was a stale-row sweep that
// returned rows to their start state: the attempt ceiling became unreachable,
// so the row cycled forever, never reached a terminal status, and was
// therefore never eligible for any retention sweep.
// 2. A PARKED step is `running` with a NULL lease, and the reclaim only ever
// touches a lease that is non-NULL and expired (the org lead's answer,
// 2026-09-02). That is what lets a GM cue wait for a human overnight without a
// sweep re-dispatching the instruction every fifteen minutes.
// A run's steps in authored order, for the phase the run is currently in.
const listForPhase = async (runId, phase) =>
(
await query(
'SELECT * FROM event_run_steps WHERE run_id = ? AND phase = ? ORDER BY seq, id',
[runId, phase],
)
).map(hydrate)
/**
* The next step of a phase that the runner may work on, or null.
*
* **Steps within a phase are strictly serial.** This returns the lowest-`seq`
* step that is not terminal, and the runner does nothing with step N+1 until N
* has finished — which is the only reading under which `core.wait` means anything
* at all, and the only one under which a cue can gate what follows it.
*
* A parked or running step is returned too, so the caller can see that the phase
* is occupied rather than concluding it is finished.
*/
const nextOpenStep = async (runId, phase) => {
const [row] = await query(
`SELECT * FROM event_run_steps
WHERE run_id = ? AND phase = ?
AND status IN ('pending','running')
ORDER BY seq, id LIMIT 1`,
[runId, phase],
)
return hydrate(row) || null
}
/**
* Take ownership of one pending step: the CAS `pending -> running`, plus a lease.
*
* `due_at` is honoured here rather than in the caller's filter so that the whole
* decision — is it mine, is it due — is one statement the database arbitrates. A
* NULL `due_at` is due now, which is what materialisation writes for every step
* that is not sitting behind a `core.wait`.
*/
async function claim(id, owner, leaseUntil, now) {
const result = await query(
`UPDATE event_run_steps
SET status = 'running', attempts = attempts + 1, claimed_by = ?, claim_expires_at = ?,
started_at = COALESCE(started_at, NOW())
WHERE id = ? AND status = 'pending' AND (due_at IS NULL OR due_at <= ?)`,
[owner, leaseUntil, id, now],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* Park a claimed step: it stays `running`, and its lease goes NULL.
*
* This is the whole mechanism behind `core.cue`. The step is genuinely in flight
* — an instruction has been posted and nothing else in the phase may proceed —
* but no process is holding it, so the reclaim must not take it back. A NULL
* lease says exactly that, and `reclaimStale` below is written to agree.
*/
const park = (id, note) =>
query(
`UPDATE event_run_steps SET claim_expires_at = NULL, last_error = ?
WHERE id = ? AND status = 'running'`,
[note ? String(note).slice(0, 500) : null, id],
)
/**
* Release a claimed step back to `pending` for a later attempt.
*
* `attempts` is untouched — it was already incremented by the claim, which is the
* only place that may. Backoff is flat rather than exponential for the reason the
* outbox's is: `due_at` is also the event's own clock, and a doubling backoff
* pushes a step arbitrarily far past the moment the event was about.
*/
const reschedule = (id, dueAt, error) =>
query(
`UPDATE event_run_steps
SET status = 'pending', due_at = ?, claimed_by = NULL, claim_expires_at = NULL, last_error = ?
WHERE id = ? AND status = 'running'`,
[dueAt, error ? String(error).slice(0, 500) : null, id],
)
/** A terminal outcome for one step: done, failed, skipped, refused or cancelled. */
const finish = (id, status, error) =>
query(
`UPDATE event_run_steps
SET status = ?, last_error = ?, finished_at = NOW(),
claimed_by = NULL, claim_expires_at = NULL
WHERE id = ? AND status = 'running'`,
[status, error ? String(error).slice(0, 500) : null, id],
)
/**
* Delay the next not-yet-started step of a phase — what `core.wait` actually does.
*
* The wait step itself completes normally; the pause is the NEXT step's `due_at`,
* owned by the runner. A `perform()` that slept would hold its claim for the
* duration and turn a five-minute pause into a five-minute lease, which is the
* one shape this must not have.
*
* Guarded on `status = 'pending'` and on the current `due_at` being sooner, so a
* re-dispatch of a wait whose ack was lost cannot push the following step further
* out a second time.
*/
const holdNext = async (runId, phase, afterSeq, dueAt) => {
const result = await query(
`UPDATE event_run_steps
SET due_at = ?
WHERE run_id = ? AND phase = ? AND seq > ? AND status = 'pending'
AND (due_at IS NULL OR due_at < ?)
ORDER BY seq LIMIT 1`,
[dueAt, runId, phase, afterSeq, dueAt],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* Recover steps whose claim outlived the process that took it.
*
* **`attempts` is not reset and the lease being NULL is not staleness.** The
* first is Engagement Phase 14's rule; the second is what makes a parked cue
* survive. A step that has already burned its attempts leaves `running` as
* `failed` rather than being handed back, and in that order — a reclaim that ran
* first would return it to `pending` and it would be retried forever.
*/
const reclaimStale = async (now, maxAttempts = 0) => {
let failed = 0
if (Number(maxAttempts) > 0) {
const gaveUp = await query(
`UPDATE event_run_steps
SET status = 'failed', last_error = 'gave up after repeated interruptions',
finished_at = NOW(), claimed_by = NULL, claim_expires_at = NULL
WHERE status = 'running'
AND claim_expires_at IS NOT NULL AND claim_expires_at < ?
AND attempts >= ?`,
[now, Math.floor(maxAttempts)],
)
failed = Number(gaveUp?.affectedRows || 0)
}
const reclaimed = await query(
`UPDATE event_run_steps
SET status = 'pending', claimed_by = NULL, claim_expires_at = NULL
WHERE status = 'running' AND claim_expires_at IS NOT NULL AND claim_expires_at < ?`,
[now],
)
return { failed, reclaimed: Number(reclaimed?.affectedRows || 0) }
}
/**
* Cancel every step of a run that has not started (Phase 3's cancel, and the
* abort_run disposition).
*
* A `running` step is deliberately left alone, parked or not: nothing can recall
* a command already sent, and a second writer on that row would race the process
* that owns it (§L).
*/
const cancelPending = async (runId) => {
const result = await query(
`UPDATE event_run_steps
SET status = 'cancelled', finished_at = NOW()
WHERE run_id = ? AND status = 'pending'`,
[runId],
)
return Number(result?.affectedRows || 0)
}
module.exports = {
listForRun,
listForPhase,
getById,
materialisePhase,
statusCounts,
idempotencyKey,
nextOpenStep,
claim,
park,
reschedule,
finish,
holdNext,
reclaimStale,
cancelPending,
}

View File

@@ -17,6 +17,12 @@ const { parseJson } = require('./eventJson')
const hydrate = (row) => row && { ...row, params: parseJson(row.params, null), rehearsal: Boolean(row.rehearsal) }
// The statuses a run never leaves. A transition INTO one of these stamps
// `ended_at` and drops the claim, and only rows in one of them are eligible for
// the log retention sweep -- Engagement Phase 14's rule, which is a bound at all
// only if every path a run can take reaches one of them.
const TERMINAL = ['completed', 'cancelled', 'failed', 'missed']
const SELECT_LIST = `
SELECT r.*, d.title AS definition_title, d.slug AS definition_slug, v.version AS version_number
FROM event_runs r
@@ -102,4 +108,256 @@ const countActiveForDefinition = async (definitionId) => {
return Number(row?.n || 0)
}
module.exports = { list, getById, materialise, findOccurrence, countActiveForDefinition }
// ── Phase 2: the claim, the transitions and the reclaim ────────────────────
//
// Everything below is the runner's, and none of it existed in Phase 1 for a
// stated reason: a half-written claim is worse than no claim, because it reads
// as protection. It is written here now, in full.
//
// **The division of labour with the unique index has not changed.** The index one
// section up is what makes "one run per occurrence per scope" TRUE; the CAS below
// decides only WHO advances an occurrence that already exists. Neither substitutes
// for the other, and this deployment being single-instance (§N4) changes the test
// rather than the design — the same two protections are what keep a tick that
// overran into the next one from advancing a run twice.
/**
* Runs the runner should look at this tick: due, and not yet terminal.
*
* It selects rather than claims — `claimStart` and `claimTick` below are one row
* at a time — so two sweepers see the same candidates and then disagree,
* harmlessly, about which of them owns each. `idx_evrun_due (status,
* scheduled_for)` is this query.
*
* `paused` is absent from the status list on purpose. A paused run is waiting on
* a human and must not be advanced by a tick; the only thing that moves it is
* Phase 3's resume control.
*/
const findDue = async (now, limit = 50) => {
const n = Math.min(Math.max(Number(limit) || 50, 1), 500)
return (
await query(
`SELECT * FROM event_runs
WHERE status IN ('scheduled','starting','running','ending')
AND scheduled_for <= ?
ORDER BY scheduled_for, id
LIMIT ${n}`,
[now],
)
).map(hydrate)
}
/**
* Take ownership of a run that has not started: the CAS `scheduled -> starting`.
*
* Verbatim the outbox claim the org lead settled over `SELECT ... FOR UPDATE
* SKIP LOCKED` — the instance the server reports `affectedRows = 1` to owns the
* row, every other sweeper gets 0 and moves on. No transaction to hold open and
* no MariaDB version floor.
*/
async function claimStart(id, owner, leaseUntil) {
const result = await query(
`UPDATE event_runs
SET status = 'starting', claimed_by = ?, claim_expires_at = ?,
started_at = COALESCE(started_at, NOW())
WHERE id = ? AND status = 'scheduled'`,
[owner, leaseUntil, id],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* Take a lease on a run already in flight, so one tick works on it at a time.
*
* Unlike `claimStart` this does not change `status` — the run is already
* `starting`, `running` or `ending`, and what is being claimed is the right to
* advance it.
*
* **A live lease is not re-enterable, not even by the process that took it**, and
* that is the whole point rather than an oversight. `setInterval` fires the next
* tick whether or not the last one has returned, so an owner-matches escape
* clause here would let one process advance one run twice at once — which is
* precisely the overrun the plan says the CAS is meant to protect against. A run
* this process still holds is a run this process is still working on; the tick
* skips it, and `releaseClaim` below is what ends that in the ordinary case.
*/
async function claimTick(id, owner, leaseUntil, now) {
const result = await query(
`UPDATE event_runs
SET claimed_by = ?, claim_expires_at = ?
WHERE id = ?
AND status IN ('starting','running','ending')
AND (claim_expires_at IS NULL OR claim_expires_at < ?)`,
[owner, leaseUntil, id, now],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* Give a still-in-flight run back, so the next tick can pick it up at once.
*
* A run left parked on a GM cue, or waiting out a `core.wait`, is not finished
* and must not carry a lease: without this the run would be unadvanceable until
* the lease expired, which would turn every wait into `max(wait, leaseMs)`.
* Scoped to `claimed_by = ?` so a process can only release its own claim.
*/
async function releaseClaim(id, owner) {
const result = await query(
'UPDATE event_runs SET claimed_by = NULL, claim_expires_at = NULL WHERE id = ? AND claimed_by = ?',
[id, owner],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* A guarded status transition: `from -> to`, and only from `from`.
*
* Every move the runner makes goes through here rather than through a bare
* UPDATE, so "did this transition actually happen" is answerable at each call
* site. A `false` is not an error — it is another worker, or this run having been
* cancelled from the admin surface between the read and the write, which is a
* race Phase 3's live controls make ordinary.
*/
async function transition(id, from, to, { phase, error, clearClaim = false } = {}) {
const sets = ['status = ?']
const args = [to]
if (phase !== undefined) {
sets.push('current_phase = ?')
args.push(phase)
}
if (error !== undefined) {
sets.push('last_error = ?')
args.push(error === null ? null : String(error).slice(0, 500))
}
if (TERMINAL.includes(to)) sets.push('ended_at = COALESCE(ended_at, NOW())')
if (clearClaim || TERMINAL.includes(to)) sets.push('claimed_by = NULL', 'claim_expires_at = NULL')
const froms = Array.isArray(from) ? from : [from]
const result = await query(
`UPDATE event_runs SET ${sets.join(', ')}
WHERE id = ? AND status IN (${froms.map(() => '?').join(',')})`,
[...args, id, ...froms],
)
return Number(result?.affectedRows || 0) === 1
}
/**
* Set health without touching status (§E).
*
* The two columns are separate because a run can be genuinely running and
* degraded at once — announcements landing, world writes parked — and one column
* cannot say both. Guarded on the current value so a tick that re-observes the
* same degradation does not restamp `updated_at`.
*/
async function setHealth(id, health) {
const result = await query('UPDATE event_runs SET health = ? WHERE id = ? AND health <> ?', [
health,
id,
health,
])
return Number(result?.affectedRows || 0) === 1
}
/**
* Runs whose start instant passed more than their own grace window ago (§E, §L).
*
* The window is per definition, so the comparison is against `grace_seconds` on
* the joined row rather than against a constant here: an event whose announcement
* gives a fifteen-minute window and one that must start on the second are the
* same query with different data.
*
* Only `scheduled` runs qualify. A run that reached `starting` has begun, and
* "began and then stalled" is a different fact from "never began" — conflating
* them would let `missed` describe a run that had already announced itself.
*/
const findMissed = async (now, limit = 100) => {
const n = Math.min(Math.max(Number(limit) || 100, 1), 500)
return (
await query(
`SELECT r.* FROM event_runs r
JOIN event_definitions d ON d.id = r.definition_id
WHERE r.status = 'scheduled'
AND r.scheduled_for + INTERVAL d.grace_seconds SECOND < ?
ORDER BY r.scheduled_for
LIMIT ${n}`,
[now],
)
).map(hydrate)
}
/**
* Is another run holding this concurrency key?
*
* The org lead's answer for a held key (2026-09-02) is to leave the run
* `scheduled` and let the grace window decide, so this is a READ rather than a
* claim: the caller holds off, logs which run holds the key, and tries again next
* tick. `idx_evrun_concurrency (concurrency_key, status)` is this query, and a
* NULL key is skipped by that index — which is right, because a definition with
* no key never contends.
*/
const concurrencyHolder = async (key, exceptRunId) => {
if (!key) return null
const [row] = await query(
`SELECT id, status, definition_id FROM event_runs
WHERE concurrency_key = ?
AND id <> ?
AND status IN ('starting','running','paused','ending')
ORDER BY id LIMIT 1`,
[key, exceptRunId],
)
return row || null
}
/**
* Recover runs whose claim outlived the process that took it.
*
* **It does not change status and it touches no counter.** All it releases is the
* lease; the run stays exactly where it was and the next tick picks it up through
* `findDue`. This is Engagement Phase 14's lesson applied one table over: a sweep
* that returned a stale row to its start state made the attempt ceiling
* unreachable, so the row cycled forever, never terminal, and therefore never
* eligible for any retention sweep.
*/
const reclaimStale = async (now) => {
const result = await query(
`UPDATE event_runs SET claimed_by = NULL, claim_expires_at = NULL
WHERE status IN ('starting','running','ending')
AND claim_expires_at IS NOT NULL
AND claim_expires_at < ?`,
[now],
)
return Number(result?.affectedRows || 0)
}
/** Terminal runs that ended before `before` — what the log retention sweep walks. */
const terminalBefore = async (before, limit = 500) => {
const n = Math.min(Math.max(Number(limit) || 500, 1), 5000)
return (
await query(
`SELECT id FROM event_runs
WHERE status IN (${TERMINAL.map(() => '?').join(',')})
AND COALESCE(ended_at, updated_at) < ?
ORDER BY id LIMIT ${n}`,
[...TERMINAL, before],
)
).map((r) => Number(r.id))
}
module.exports = {
list,
getById,
materialise,
findOccurrence,
countActiveForDefinition,
findDue,
findMissed,
claimStart,
claimTick,
releaseClaim,
transition,
setHealth,
concurrencyHolder,
reclaimStale,
terminalBefore,
TERMINAL,
}

View File

@@ -15,6 +15,7 @@ const engagementRetentionPrune = require('./utils/engagementRetentionPrune')
const teamForumUploadSweep = require('./utils/teamForumUploadSweep')
const teamDigestWorker = require('./utils/teamDigestWorker')
const engagementWorker = require('./utils/engagementWorker')
const eventRunner = require('./utils/eventRunner')
const { ensureSchema, close } = require('./utils/db')
const { seedDefaults, createInitialAdminFromEnv } = require('../db/seed')
const settings = require('./model/settings/settings.model')
@@ -169,6 +170,11 @@ async function start() {
// enables a rule: core seeds none and `enabled` defaults to 0.
engagementWorker.start()
// Advance scheduled events (EVENTS.md §E). Materialise, advance, drain, and the
// run-log retention sweep. No-op until an admin publishes a definition and
// starts a run: core ships no event definitions.
eventRunner.start()
setupShutdown(server, internalServer)
}
@@ -192,6 +198,7 @@ function setupShutdown(server, internalServer) {
teamForumUploadSweep.stop() // stop the forum upload sweep
teamDigestWorker.stop() // stop the Team forum digest timer
engagementWorker.stop() // stop the engagement outbox worker
eventRunner.stop() // stop the event runner
server.close(() => log.info('http server closed'))
if (internalServer) internalServer.close(() => log.info('internal http server closed'))
try {

View File

@@ -0,0 +1,510 @@
// ── The event runner ───────────────────────────────────────────────────────
//
// EVENTS.md §E, and Phase 2 of EVENTS_PLAN.md. The eighth poller: same
// `setInterval` + `unref()` + `stop()` shape as `announceWorker`, the three Team
// sweepers and `engagementWorker`, wired into `server.js` beside them. In the
// **website** process rather than the bot, which cannot load module code.
//
// Its tick does four things, in this order:
//
// 1. **reclaim** — release leases whose holder died, never touching `attempts`
// 2. **materialise** — sweep occurrences past their grace window into `missed`
// 3. **advance** — claim each due run and move it through its phases
// 4. **prune** — the `event_run_log` retention sweep, on its own long clock
//
// **What "materialise" means in this phase.** §E's tick materialises due
// occurrences from a recurrence; the spec validator accepts `kind: 'manual'`
// alone until Phase 4, so there is no recurrence to expand and the only
// occurrences that exist are the ones an admin created. What that leaves for this
// leg is the half that is already real and already needed: the grace window. A
// run whose instant passed while the process was down does not start late and
// silently — it becomes `missed`, which is terminal and which a human can see
// (§L). Phase 4 adds the expansion above it.
//
// **Two properties this file must not lose**, both already paid for once on this
// codebase:
//
// - **A reclaim never resets `attempts`.** Engagement Phase 14's defect: a sweep
// that returned every stale row to its start state made the attempt ceiling
// unreachable, so the row cycled forever, never terminal, and therefore never
// eligible for any retention sweep.
// - **The unique index, not the claim, is what prevents a double run.** The claim
// decides *who* advances an occurrence; `uq_evrun_occurrence` is what stops two
// of them existing. Neither substitutes for the other.
//
// §N4 settled this deployment as single-instance, so the `--scale app=2` rig the
// engagement workstream used is deliberately not built. Every claim path is here
// exactly as §E specifies anyway, because the multi-instance case is not what
// they are for on a single container: the CAS is what protects a tick that
// overran into the next one, and the lease and its reclaim are what recover a
// step whose process died mid-dispatch. Both happen with one app.
const os = require('os')
const runsDb = require('../model/events/eventRuns.db')
const stepsDb = require('../model/events/eventRunSteps.db')
const logDb = require('../model/events/eventRunLog.db')
const versionsDb = require('../model/events/eventVersions.db')
const registries = require('../modules/registries')
const { dispatchStep } = require('../events/dispatch')
const log = require('./logger')('event-runner')
const POLL_MS = Number(process.env.EVENT_POLL_MS) || 15_000
// How many runs one sweep looks at, and how many steps it will drain from one
// run. Bounds rather than targets: the tick runs again in POLL_MS, and an
// unbounded batch is how a backlog turns one tick into a stall. The step bound
// also caps how long one run can hold the tick, which is what keeps a
// forty-step phase from starving every other run on the deployment.
const RUN_BATCH = Number(process.env.EVENT_RUN_BATCH) || 50
const STEPS_PER_TICK = Number(process.env.EVENT_STEPS_PER_TICK) || 25
// A step's retries (org lead, 2026-09-02). §L specifies `retry(n) -> skip |
// pause | abort_run` and names no `n`; it lives here, as one number an operator
// can change, rather than in a column no authoring surface would ever show.
//
// Flat backoff rather than exponential, for the reason the outbox's is flat:
// `due_at` is also the event's own clock, and a doubling backoff pushes a step
// arbitrarily far past the moment the event was about.
const MAX_ATTEMPTS = Number(process.env.EVENT_STEP_MAX_ATTEMPTS) || 3
const RETRY_MS = Number(process.env.EVENT_STEP_RETRY_MS) || 60_000
// How long a claim may look alive before the reclaim takes it back. The run
// lease has to outlast a whole tick's work on one run; a step's is computed from
// its own action's `budgetMs` (see `leaseFor`), because a registry that lets an
// action declare an hour would otherwise have its steps reclaimed and
// re-dispatched fifty-nine minutes before they answered.
const RUN_LEASE_MS = Number(process.env.EVENT_RUN_LEASE_MS) || 15 * 60 * 1000
const STEP_LEASE_MARGIN_MS = 60_000
// The log retention horizon, and how often the sweep runs. It is folded into
// this tick rather than given a ninth timer because it shares the tick's only
// real dependency — a database — and a sweep that runs four times a day does not
// need an interval of its own. Only TERMINAL runs are ever eligible, which is the
// rule Engagement Phase 14 arrived at.
const LOG_RETENTION_DAYS = Number(process.env.EVENT_LOG_RETENTION_DAYS) || 90
const PRUNE_EVERY_MS = 6 * 60 * 60 * 1000
// Who this process is, for `claimed_by`. Host and pid, so a stranded claim in the
// table names the thing that stranded it.
const OWNER = `${os.hostname()}:${process.pid}`.slice(0, 64)
const RUN_TERMINAL = runsDb.TERMINAL
/** The lease a step's dispatch gets: its action's own budget, plus a margin. */
function leaseFor(step, now) {
const action = registries.eventAction(step.action_id)
const budget = action?.budgetMs || 10_000
return new Date(now.getTime() + budget + STEP_LEASE_MARGIN_MS)
}
// ── The disposition of a step that has run out of road ─────────────────────
//
// **All three dispositions write the step `failed`.** `on_failure` says what
// happens to the RUN, not what happened to the step, and a step that was
// attempted three times and never worked is `failed` under every one of them.
// `skipped` is reserved for a step a human skipped from the run console (Phase
// 3) — a status that meant both "nobody ran this" and "this failed and we moved
// on" would make the run console's summary line unreadable.
async function applyFailure(run, step, error) {
await stepsDb.finish(step.id, 'failed', error)
await logDb.write({
runId: run.id,
stepId: step.id,
kind: 'step.status',
phase: step.phase,
detail: { to: 'failed', action: step.action_id, attempts: step.attempts + 1, onFailure: step.on_failure, error },
})
// A run that lost a step is degraded whatever happens next. Health is not
// status (§E): a run can be genuinely running and degraded at once, and the
// admin surface needs to say so without claiming the run stopped.
if (await runsDb.setHealth(run.id, 'degraded')) {
await logDb.write({ runId: run.id, kind: 'run.health', detail: { to: 'degraded', because: step.action_id } })
}
if (step.on_failure === 'abort_run') {
// §L: the disposition for `irreversible`. Nothing further is dispatched, and
// the pending steps are cancelled rather than left looking due forever.
const cancelled = await stepsDb.cancelPending(run.id)
await runsDb.transition(run.id, ['starting', 'running', 'ending'], 'failed', { error })
await logDb.write({
runId: run.id,
kind: 'run.status',
phase: step.phase,
detail: { to: 'failed', because: step.action_id, cancelledSteps: cancelled },
})
return 'stop'
}
if (step.on_failure === 'pause') {
// The default for `change`, and the right one when the world is half-altered:
// stop advancing and wait for a human. `paused` is excluded from `findDue`,
// so nothing here picks it up again — Phase 3's resume control is the only
// thing that moves it.
await runsDb.transition(run.id, ['starting', 'running'], 'paused', { error })
await logDb.write({
runId: run.id,
kind: 'run.status',
phase: step.phase,
detail: { to: 'paused', because: step.action_id },
})
return 'stop'
}
// 'skip': the run carries on, degraded, and the failure is on the record.
return 'continue'
}
/**
* Claim one step, dispatch it, and record what came back.
*
* Answers `'continue'` (the phase may proceed), `'stop'` (it may not, for now or
* ever) or `'taken'` (somebody else claimed it first).
*/
async function drainStep(run, step, now, carry = {}) {
if (!(await stepsDb.claim(step.id, OWNER, leaseFor(step, now), now))) return 'taken'
const result = await dispatchStep(step, { run })
if (result.actionVersionDrift) {
await logDb.write({
runId: run.id,
stepId: step.id,
kind: 'step.status',
phase: step.phase,
detail: { action: step.action_id, versionDrift: result.actionVersionDrift },
})
}
if (result.outcome === 'parked') {
// The GM cue. The step stays `running` with a NULL lease: genuinely in
// flight, nothing holding it, so the stale reclaim passes it by and a cue
// posted on Friday is still waiting on Monday. Phase 3's confirm control is
// what ends it.
await stepsDb.park(step.id, null)
await logDb.write({
runId: run.id,
stepId: step.id,
kind: 'step.parked',
phase: step.phase,
detail: { action: step.action_id, params: step.params },
})
return 'stop'
}
if (result.outcome === 'done') {
await stepsDb.finish(step.id, 'done', null)
if (result.holdSeconds > 0) {
// `core.wait`, and any module action that answers `holdFor`. The pause is
// the NEXT step's `due_at` and it is set here, by the runner, because a
// `perform()` that slept would hold its claim for the duration.
const until = new Date(now.getTime() + result.holdSeconds * 1000)
if (!(await stepsDb.holdNext(run.id, step.phase, step.seq, until))) {
// **Nothing after it in this phase**, which is the case a wait written as
// the last step of a phase produces. Dropping the hold here would make
// "announce, wait five minutes, then the next phase" start the next phase
// at once — a wait that silently meant nothing. The later phase's steps do
// not exist yet, so the instant is carried out to `advanceRun` and applied
// when they are materialised.
carry.holdUntil = until
}
}
await logDb.write({
runId: run.id,
stepId: step.id,
kind: 'step.status',
phase: step.phase,
detail: { to: 'done', action: step.action_id, holdSeconds: result.holdSeconds || 0 },
})
return 'continue'
}
if (result.outcome === 'retry' && step.attempts + 1 < MAX_ATTEMPTS) {
await stepsDb.reschedule(step.id, new Date(now.getTime() + RETRY_MS), result.error)
await logDb.write({
runId: run.id,
stepId: step.id,
kind: 'step.retry',
phase: step.phase,
detail: { action: step.action_id, attempt: step.attempts + 1, of: MAX_ATTEMPTS, error: result.error },
})
// Degraded from the FIRST retry, not from the eventual failure. An event
// whose announcements are landing on the second attempt is having trouble
// now, and that is when an operator wants to know.
if (await runsDb.setHealth(run.id, 'degraded')) {
await logDb.write({ runId: run.id, kind: 'run.health', detail: { to: 'degraded', because: step.action_id } })
}
return 'stop'
}
// Terminal, or transient with the attempts spent. Same disposition either way:
// §L's `retry(n) -> ...` has arrived at the arrow.
return applyFailure(run, step, result.error)
}
/**
* Advance one claimed run as far as it will go this tick.
*
* The loop is bounded by `STEPS_PER_TICK` and exits on the first thing it cannot
* get past — a parked step, a step whose `due_at` is in the future, a step
* somebody else holds, or a phase that is not finished.
*/
async function advanceRun(run, now) {
const version = await versionsDb.getById(run.version_id)
const phases = version?.spec?.phases
if (!Array.isArray(phases) || !phases.length) {
// The pinned version is unreadable. `version_id`'s foreign key RESTRICTs
// precisely so this cannot be a deleted row, so it is corruption rather than
// an ordinary race — terminal, named, and not retried.
await runsDb.transition(run.id, ['scheduled', 'starting', 'running', 'ending'], 'failed', {
error: 'the pinned version has no phases',
})
await logDb.write({ runId: run.id, kind: 'run.status', detail: { to: 'failed', because: 'pinned version has no phases' } })
return 'failed'
}
let phaseKey = run.current_phase
if (run.status === 'starting') {
// Entering the first phase. `materialisePhase` is INSERT IGNORE against
// `uq_evstep_slot`, so doing it again over the rows Phase 1's `create()`
// already wrote is a no-op — which is what makes recovery from a process that
// died between the claim and here uneventful.
const first = phases[0]
await stepsDb.materialisePhase(run.id, first.key, first.steps || [])
if (!(await runsDb.transition(run.id, 'starting', 'running', { phase: first.key }))) return 'taken'
phaseKey = first.key
await logDb.write({ runId: run.id, kind: 'run.status', phase: first.key, detail: { from: 'starting', to: 'running' } })
}
if (run.status === 'ending') {
// A run that reached the wind-down and then lost its process. Phase 8 puts
// cleanup here; until then `ending` is a state a run passes through rather
// than one it does work in, and completing it is the whole recovery.
await runsDb.transition(run.id, 'ending', 'completed')
await logDb.write({ runId: run.id, kind: 'run.status', detail: { from: 'ending', to: 'completed' } })
return 'completed'
}
// A hold a `core.wait` could not place because nothing followed it in its own
// phase. It crosses the phase boundary with the run rather than being dropped.
const carry = {}
for (let n = 0; n < STEPS_PER_TICK; n++) {
const phaseIndex = phases.findIndex((p) => p.key === phaseKey)
if (phaseIndex < 0) {
await runsDb.transition(run.id, ['running'], 'failed', { error: `phase "${phaseKey}" is not in the pinned version` })
await logDb.write({ runId: run.id, kind: 'run.status', detail: { to: 'failed', because: `unknown phase "${phaseKey}"` } })
return 'failed'
}
const step = await stepsDb.nextOpenStep(run.id, phaseKey)
if (step && step.status === 'running') return 'in-flight' // parked, or somebody's dispatch
if (step && step.due_at && new Date(step.due_at) > now) return 'waiting' // behind a core.wait
if (step) {
const outcome = await drainStep({ ...run, current_phase: phaseKey }, step, now, carry)
if (outcome === 'continue') continue
return outcome === 'taken' ? 'taken' : 'stopped'
}
// Every step of this phase is terminal.
await logDb.write({ runId: run.id, kind: 'phase.completed', phase: phaseKey, detail: { index: phaseIndex } })
const next = phases[phaseIndex + 1]
if (!next) {
// §E's `ending` exists for the reason `sending` does in the outbox — it is
// what a claim sets — so the run passes through it even though Phase 2 has
// no cleanup to do there. Phase 8 is what gives it work.
if (!(await runsDb.transition(run.id, 'running', 'ending'))) return 'taken'
await logDb.write({ runId: run.id, kind: 'run.status', phase: phaseKey, detail: { from: 'running', to: 'ending' } })
await runsDb.transition(run.id, 'ending', 'completed')
await logDb.write({ runId: run.id, kind: 'run.status', detail: { from: 'ending', to: 'completed' } })
return 'completed'
}
await stepsDb.materialisePhase(run.id, next.key, next.steps || [])
if (carry.holdUntil) {
// `seq > -1` is the first step of the phase just created. Applied after
// materialisation because that is the first moment there is a row to hold.
await stepsDb.holdNext(run.id, next.key, -1, carry.holdUntil)
carry.holdUntil = null
}
// `running -> running` is not a no-op: it is a guarded write of
// `current_phase` that fails if the run stopped being `running` underneath
// this tick, which is what a cancel from the admin surface looks like.
if (!(await runsDb.transition(run.id, 'running', 'running', { phase: next.key }))) return 'taken'
phaseKey = next.key
await logDb.write({ runId: run.id, kind: 'phase.entered', phase: next.key, detail: { steps: (next.steps || []).length } })
}
return 'bounded' // more to do; the next tick picks it up
}
/** Claim one due run, work it, and hand the lease back if it is still in flight. */
async function processRun(run, now = new Date()) {
if (run.status === 'scheduled') {
const holder = await runsDb.concurrencyHolder(run.concurrency_key, run.id)
if (holder) {
// The org lead's answer for a held key (2026-09-02): hold at `scheduled`
// and let the grace window decide. Nothing is destroyed, nothing starts
// silently late, and if the holder outlasts the window the missed sweep
// makes this run terminal and visible.
//
// Logged only when the reason CHANGES. A blocked run is re-examined every
// tick, and a line per tick for the length of a grace window would bury the
// one line that matters under a thousand identical ones.
const message = `held: run ${holder.id} has concurrency key "${run.concurrency_key}"`
if (run.last_error !== message) {
await runsDb.transition(run.id, 'scheduled', 'scheduled', { error: message })
await logDb.write({
runId: run.id,
kind: 'run.blocked',
detail: { concurrencyKey: run.concurrency_key, heldBy: holder.id, holderStatus: holder.status },
})
}
return 'blocked'
}
if (!(await runsDb.claimStart(run.id, OWNER, new Date(now.getTime() + RUN_LEASE_MS)))) return 'taken'
await logDb.write({ runId: run.id, kind: 'run.status', detail: { from: 'scheduled', to: 'starting' } })
run = { ...run, status: 'starting' }
} else if (!(await runsDb.claimTick(run.id, OWNER, new Date(now.getTime() + RUN_LEASE_MS), now))) {
return 'taken'
}
try {
const outcome = await advanceRun(run, now)
// A run that is not finished must not keep its lease: it would be
// unadvanceable until the lease expired, which would turn every `core.wait`
// into `max(wait, RUN_LEASE_MS)`. A terminal transition already cleared it.
if (!['completed', 'failed'].includes(outcome)) await runsDb.releaseClaim(run.id, OWNER)
return outcome
} catch (err) {
await runsDb.releaseClaim(run.id, OWNER)
throw err
}
}
/** Occurrences that passed their own grace window while nothing was running (§L). */
async function sweepMissed(now) {
const missed = await runsDb.findMissed(now)
let n = 0
for (const run of missed) {
if (await runsDb.transition(run.id, 'scheduled', 'missed', { error: 'the grace window passed' })) {
await stepsDb.cancelPending(run.id)
await logDb.write({
runId: run.id,
kind: 'run.status',
detail: { to: 'missed', scheduledFor: run.scheduled_for },
})
n += 1
}
}
return n
}
let lastPruneAt = 0
async function prune(now) {
if (now.getTime() - lastPruneAt < PRUNE_EVERY_MS) return 0
lastPruneAt = now.getTime()
const before = new Date(now.getTime() - LOG_RETENTION_DAYS * 24 * 60 * 60 * 1000)
const deleted = await logDb.pruneTerminal(before)
if (deleted) log.info('event run log pruned', { deleted, before, retentionDays: LOG_RETENTION_DAYS })
return deleted
}
async function tick(now = new Date()) {
try {
await runsDb.reclaimStale(now)
await stepsDb.reclaimStale(now, MAX_ATTEMPTS)
} catch (err) {
log.error('failed to reclaim stale claims', { message: err.message })
}
try {
await sweepMissed(now)
} catch (err) {
log.error('missed sweep failed', { message: err.message })
}
let due
try {
due = await runsDb.findDue(now, RUN_BATCH)
} catch (err) {
log.error('failed to load due runs', { message: err.message })
return
}
const counts = {}
for (const run of due || []) {
try {
const outcome = await processRun(run, now)
counts[outcome] = (counts[outcome] || 0) + 1
} catch (err) {
log.error('run failed', { run: run.id, message: err.message })
}
}
if (due && due.length) log.info('event runs swept', { due: due.length, ...counts })
try {
await prune(now)
} catch (err) {
log.error('log prune failed', { message: err.message })
}
}
let timer = null
// Guards against this process running two ticks over the same runs at once.
// `setInterval` fires whether or not the last callback returned, and the CAS
// alone does not cover it now that a live lease is not re-enterable by its own
// owner — a second tick would simply find every run claimed and do nothing
// useful, one query at a time, for as long as the first one ran.
let ticking = false
function start() {
if (timer) return timer
timer = setInterval(() => {
if (ticking) {
log.warn('event tick still running; skipping this interval')
return
}
ticking = true
tick()
.catch((err) => log.error('event tick failed', { message: err.message }))
.finally(() => {
ticking = false
})
}, POLL_MS)
if (timer.unref) timer.unref() // don't keep the event loop alive (tests, shutdown)
log.info('event runner started', { pollMs: POLL_MS, owner: OWNER, maxAttempts: MAX_ATTEMPTS })
return timer
}
function stop() {
if (timer) {
clearInterval(timer)
timer = null
}
}
module.exports = {
start,
stop,
tick,
processRun,
advanceRun,
drainStep,
sweepMissed,
prune,
OWNER,
POLL_MS,
MAX_ATTEMPTS,
RETRY_MS,
RUN_LEASE_MS,
LOG_RETENTION_DAYS,
RUN_TERMINAL,
}

View File

@@ -63,17 +63,70 @@ test('the catalog carries no callable', () => {
assert.equal(typeof registries.eventAction('core.wait').perform, 'function')
})
test('core placeholders refuse rather than claiming success', async () => {
// Phase 1 declares; Phase 2 dispatches. The placeholder's answer matters
// because `ok: true` on an action that did nothing is a recorded world change
// that did not occur — the one wrong answer a stub can give.
test('core.announce refuses an unregistered leg terminally, and never claims success', async () => {
// Phase 1's version of this test asserted that all three core actions REFUSED,
// because none of them was wired yet. Phase 2 gave them real bodies, so what
// survives is the half that was never about the placeholder: `ok: true` on an
// action that did nothing is a recorded world change that did not occur.
//
// `core.announce` is the one that can still legitimately refuse. A leg nobody
// registers will not appear between two attempts a minute apart, so the answer
// is terminal rather than transient — a human has to fix it.
registries.registerCore()
for (const id of ['core.announce', 'core.wait', 'core.cue']) {
const answer = await registries.eventAction(id).perform({})
assert.equal(answer.ok, false)
assert.equal(answer.retry, false)
assert.match(answer.error, new RegExp(id.replace('.', '\\.')))
const answer = await registries.eventAction('core.announce').perform({
params: { leg: 'nowhere', body: 'hello' },
})
assert.equal(answer.ok, false)
assert.equal(answer.retry, false)
assert.match(answer.error, /nowhere/)
})
test('a dry run validates and reports, but dispatches nothing', async () => {
// §I's dry run: `verify === true` means validate and report, change nothing.
// `core.announce` is the only core action with an outside effect to suppress.
//
// **The leg is resolved BEFORE `verify` is honoured, and that ordering is the
// point rather than an oversight.** A dry run exists to report what would
// happen, and "this step names a leg nobody registers" is the most useful thing
// it can find. Answering `ok: true` first would make the dry run pass on
// exactly the definition that cannot work.
registries.registerCore()
const announce = registries.eventAction('core.announce')
const bad = await announce.perform({ params: { leg: 'nowhere', body: 'hello' }, verify: true })
assert.equal(bad.ok, false, 'a dry run must surface a leg that does not exist')
// A registered leg: reported good, and its transport never touched.
const leg = registries.announceLeg('discord')
const dispatch = leg.dispatch
let dispatched = 0
leg.dispatch = async () => {
dispatched += 1
return { ok: true }
}
try {
const good = await announce.perform({ params: { leg: 'discord', body: 'hello' }, verify: true })
assert.equal(good.ok, true)
assert.equal(dispatched, 0, 'a dry run sends nothing')
} finally {
leg.dispatch = dispatch
}
})
test('core.wait defers the next step rather than sleeping, and core.cue parks', async () => {
// Both answer through ordinary envelope members, which is what lets the runner
// honour them without knowing what either action is. A `perform` that slept
// would hold its claim for the duration and turn a five-minute pause into a
// five-minute lease.
registries.registerCore()
assert.deepEqual(await registries.eventAction('core.wait').perform({ params: { seconds: 300 } }), {
ok: true,
holdFor: 300,
})
assert.deepEqual(await registries.eventAction('core.cue').perform({ params: {} }), {
ok: true,
await: 'human',
})
})
test('an action must be namespaced to its owner, and the holder is named', () => {

View File

@@ -0,0 +1,689 @@
// ── The event runner (EVENTS_PLAN.md Phase 2) ──────────────────────────────
//
// The phase's shipped claim, first: **a manually started event that broadcasts,
// waits, and completes.** Then the properties around it that are not behaviour
// so much as promises — the ones §E and §L make, and the two this codebase has
// already paid for once:
//
// • a reclaim never resets `attempts` (Engagement Phase 14's defect)
// • a parked GM cue is not stale, however long it waits
// • no shape a failure can take reads as success (§F)
// • a step naming an unregistered action fails terminal with the module named
// and degrades the run — never a silent skip (§L)
// • the three `on_failure` dispositions do three different things to the RUN
// • the idempotency key does not vary by attempt (§E)
//
// **The three tables are stubbed at the `.db` layer** and the runner's own logic
// runs for real against them — the shape `engagementEngine.test.js` uses. What a
// stub cannot prove is the raw SQL whose correctness IS a server contract: the
// two CAS claims, the lease reclaim's two-statement order, and `holdNext`'s
// guard. Those run against a real MariaDB in `eventRunnerSql.test.js`, which
// skips when there is none. A stub reproduces the reading, not the server.
//
// Point the DB at a closed port before requiring anything: the registries reach
// utils/discordAnnounce, which builds the pool at require time.
process.env.DB_HOST = '127.0.0.1'
process.env.DB_PORT = '59999'
const { test, beforeEach, afterEach, after } = require('node:test')
const assert = require('node:assert/strict')
const registries = require('../src/modules/registries')
const runner = require('../src/utils/eventRunner')
const { classify } = require('../src/events/dispatch')
const runsDb = require('../src/model/events/eventRuns.db')
const stepsDb = require('../src/model/events/eventRunSteps.db')
const logDb = require('../src/model/events/eventRunLog.db')
const versionsDb = require('../src/model/events/eventVersions.db')
const db = require('../src/utils/db')
after(() => db.close())
const T0 = new Date('2026-09-02T12:00:00Z')
const later = (ms) => new Date(T0.getTime() + ms)
// ── In-memory stand-ins for the three tables ───────────────────────────────
let store
const originals = {}
for (const [name, mod] of [['runsDb', runsDb], ['stepsDb', stepsDb], ['logDb', logDb], ['versionsDb', versionsDb]]) {
originals[name] = { mod, fns: { ...mod } }
}
const restoreOriginals = () => {
for (const { mod, fns } of Object.values(originals)) Object.assign(mod, fns)
}
const TERMINAL_RUN = ['completed', 'cancelled', 'failed', 'missed']
const clone = (o) => JSON.parse(JSON.stringify(o, (k, v) => v))
function installStubs() {
store = {
runs: new Map(),
steps: new Map(),
log: [],
versions: new Map(),
definitions: new Map(),
nextStepId: 1,
}
// Snapshots, not live references. A SQL SELECT hands back a copy, and the
// runner reads `step.attempts` as the value BEFORE its own claim incremented
// it — returning references here would make the retry budget off by one in the
// stub only, which is exactly the class of thing a stub must not invent.
const snapRun = (r) => ({ ...r })
const snapStep = (s) => ({ ...s, params: { ...(s.params || {}) } })
runsDb.findDue = async (now) =>
[...store.runs.values()]
.filter((r) => ['scheduled', 'starting', 'running', 'ending'].includes(r.status) && r.scheduled_for <= now)
.sort((a, b) => a.scheduled_for - b.scheduled_for || a.id - b.id)
.map(snapRun)
runsDb.findMissed = async (now) =>
[...store.runs.values()]
.filter((r) => {
const grace = store.definitions.get(r.definition_id)?.grace_seconds ?? 900
return r.status === 'scheduled' && r.scheduled_for.getTime() + grace * 1000 < now.getTime()
})
.map(snapRun)
runsDb.claimStart = async (id, owner, lease) => {
const r = store.runs.get(id)
if (!r || r.status !== 'scheduled') return false
Object.assign(r, { status: 'starting', claimed_by: owner, claim_expires_at: lease, started_at: r.started_at || T0 })
return true
}
runsDb.claimTick = async (id, owner, lease, now) => {
const r = store.runs.get(id)
if (!r || !['starting', 'running', 'ending'].includes(r.status)) return false
// No owner-matches escape: a live lease is not re-enterable, not even by the
// process that took it. The stub agrees with the statement on purpose.
if (r.claim_expires_at && r.claim_expires_at >= now) return false
Object.assign(r, { claimed_by: owner, claim_expires_at: lease })
return true
}
runsDb.releaseClaim = async (id, owner) => {
const r = store.runs.get(id)
if (!r || r.claimed_by !== owner) return false
Object.assign(r, { claimed_by: null, claim_expires_at: null })
return true
}
runsDb.transition = async (id, from, to, opts = {}) => {
const r = store.runs.get(id)
const froms = Array.isArray(from) ? from : [from]
if (!r || !froms.includes(r.status)) return false
r.status = to
if (opts.phase !== undefined) r.current_phase = opts.phase
if (opts.error !== undefined) r.last_error = opts.error
if (TERMINAL_RUN.includes(to)) {
r.ended_at = r.ended_at || T0
r.claimed_by = null
r.claim_expires_at = null
} else if (opts.clearClaim) {
r.claimed_by = null
r.claim_expires_at = null
}
return true
}
runsDb.setHealth = async (id, health) => {
const r = store.runs.get(id)
if (!r || r.health === health) return false
r.health = health
return true
}
runsDb.concurrencyHolder = async (key, exceptId) => {
if (!key) return null
const held = [...store.runs.values()].find(
(r) => r.concurrency_key === key && r.id !== exceptId && ['starting', 'running', 'paused', 'ending'].includes(r.status),
)
return held ? { id: held.id, status: held.status, definition_id: held.definition_id } : null
}
runsDb.reclaimStale = async (now) => {
let n = 0
for (const r of store.runs.values()) {
if (['starting', 'running', 'ending'].includes(r.status) && r.claim_expires_at && r.claim_expires_at < now) {
r.claimed_by = null
r.claim_expires_at = null
n += 1
}
}
return n
}
stepsDb.materialisePhase = async (runId, phase, steps) => {
steps.forEach((s, i) => {
// INSERT IGNORE against uq_evstep_slot (run_id, phase, seq).
const exists = [...store.steps.values()].find((x) => x.run_id === runId && x.phase === phase && x.seq === i)
if (exists) return
const id = store.nextStepId++
store.steps.set(id, {
id,
run_id: runId,
phase,
seq: i,
action_id: s.actionId,
params: s.params || {},
action_version: s.actionVersion || 1,
status: 'pending',
due_at: null,
attempts: 0,
on_failure: s.onFailure || 'pause',
idempotency_key: stepsDb.idempotencyKey(runId, id),
claimed_by: null,
claim_expires_at: null,
last_error: null,
})
})
return [...store.steps.values()].filter((s) => s.run_id === runId).map(snapStep)
}
stepsDb.nextOpenStep = async (runId, phase) => {
const s = [...store.steps.values()]
.filter((x) => x.run_id === runId && x.phase === phase && ['pending', 'running'].includes(x.status))
.sort((a, b) => a.seq - b.seq || a.id - b.id)[0]
return s ? snapStep(s) : null
}
stepsDb.claim = async (id, owner, lease, now) => {
const s = store.steps.get(id)
if (!s || s.status !== 'pending') return false
if (s.due_at && s.due_at > now) return false
Object.assign(s, { status: 'running', attempts: s.attempts + 1, claimed_by: owner, claim_expires_at: lease })
return true
}
stepsDb.park = async (id) => {
const s = store.steps.get(id)
if (s && s.status === 'running') s.claim_expires_at = null
}
stepsDb.reschedule = async (id, dueAt, error) => {
const s = store.steps.get(id)
if (!s || s.status !== 'running') return
// `attempts` is untouched: the claim already incremented it, and nothing else
// may. This is the stub agreeing with the statement, not with the runner.
Object.assign(s, { status: 'pending', due_at: dueAt, claimed_by: null, claim_expires_at: null, last_error: error })
}
stepsDb.finish = async (id, status, error) => {
const s = store.steps.get(id)
if (!s || s.status !== 'running') return
Object.assign(s, { status, last_error: error, claimed_by: null, claim_expires_at: null, finished_at: T0 })
}
stepsDb.holdNext = async (runId, phase, afterSeq, dueAt) => {
const s = [...store.steps.values()]
.filter((x) => x.run_id === runId && x.phase === phase && x.seq > afterSeq && x.status === 'pending')
.filter((x) => !x.due_at || x.due_at < dueAt)
.sort((a, b) => a.seq - b.seq)[0]
if (!s) return false
s.due_at = dueAt
return true
}
stepsDb.reclaimStale = async (now, maxAttempts = 0) => {
let failed = 0
let reclaimed = 0
// Give up first, reclaim second — the order the statement uses, and the
// reason `MAX_ATTEMPTS` is reachable at all.
for (const s of store.steps.values()) {
if (s.status === 'running' && s.claim_expires_at && s.claim_expires_at < now && maxAttempts > 0 && s.attempts >= maxAttempts) {
Object.assign(s, { status: 'failed', last_error: 'gave up after repeated interruptions', claimed_by: null, claim_expires_at: null })
failed += 1
}
}
for (const s of store.steps.values()) {
if (s.status === 'running' && s.claim_expires_at && s.claim_expires_at < now) {
// NOT reset: attempts survives the reclaim.
Object.assign(s, { status: 'pending', claimed_by: null, claim_expires_at: null })
reclaimed += 1
}
}
return { failed, reclaimed }
}
stepsDb.cancelPending = async (runId) => {
let n = 0
for (const s of store.steps.values()) {
if (s.run_id === runId && s.status === 'pending') {
s.status = 'cancelled'
n += 1
}
}
return n
}
stepsDb.listForRun = async (runId) =>
[...store.steps.values()].filter((s) => s.run_id === runId).sort((a, b) => a.seq - b.seq).map(snapStep)
logDb.write = async (line) => {
store.log.push(line)
return true
}
logDb.pruneTerminal = async () => 0
versionsDb.getById = async (id) => store.versions.get(id) || null
}
// ── Fixtures ───────────────────────────────────────────────────────────────
let nextRunId = 1
function seedRun(phases, { scheduledFor = T0, graceSeconds = 900, concurrencyKey = null, status = 'scheduled' } = {}) {
const id = nextRunId++
store.definitions.set(id, { id, grace_seconds: graceSeconds })
store.versions.set(id, { id, spec: { schedule: { kind: 'manual' }, phases } })
store.runs.set(id, {
id,
definition_id: id,
version_id: id,
scope: '',
status,
health: 'ok',
cleanup_status: 'not_required',
current_phase: null,
scheduled_for: scheduledFor,
concurrency_key: concurrencyKey,
params: null,
rehearsal: 0,
claimed_by: null,
claim_expires_at: null,
last_error: null,
started_at: null,
ended_at: null,
})
// Phase 1's `create()` materialises the FIRST phase at creation rather than at
// start, so a seeded run has to as well — otherwise every test here would be
// exercising a shape the admin route cannot produce.
const first = phases[0]
if (first) void stepsDb.materialisePhase(id, first.key, first.steps || [])
return id
}
const step = (actionId, params = {}, onFailure = 'skip') => ({ actionId, params, onFailure, actionVersion: 1 })
const run = (id) => store.runs.get(id)
const stepsOf = (id) => [...store.steps.values()].filter((s) => s.run_id === id).sort((a, b) => a.seq - b.seq)
const kinds = (id) => store.log.filter((l) => l.runId === id).map((l) => l.kind)
// A registered test action whose behaviour the test dictates.
let scripted
beforeEach(() => {
registries._reset()
installStubs()
nextRunId = 1
scripted = {}
})
afterEach(() => {
restoreOriginals()
registries._reset()
})
/** Register actions the way a module does, through the real staging area. */
const register = (entries, owner = 'test') => {
const api = registries.stage(owner)
api.registerEventActions(entries)
registries.apply(api.staged)
}
// ── Registering actions the tests drive ────────────────────────────────────
//
// Registered through the real registry rather than by stubbing `eventAction`,
// because the shape check at registration is part of what the runner relies on:
// an action that would not register is not one the runner has to survive.
const scriptedAction = (id, extra = {}) => ({
id,
label: id,
risk: 'notify',
reversible: 'none',
version: 1,
budgetMs: 1000,
params: [],
perform: async (envelope) => {
;(scripted[id] ||= { calls: [] }).calls.push(envelope)
const answer = scripted[id].answers?.shift() ?? scripted[id].answer
if (typeof answer === 'function') return answer(envelope)
return answer ?? { ok: true }
},
...extra,
})
test('a manually started event announces, waits and completes', async () => {
register([scriptedAction('test.announce'), scriptedAction('test.wait')])
scripted['test.wait'] = { calls: [], answer: { ok: true, holdFor: 300 } }
const id = seedRun([
{ key: 'main', label: 'Main', steps: [step('test.announce'), step('test.wait'), step('test.announce')] },
])
// Tick one: announce, then wait, then stop against the held third step.
await runner.tick(T0)
let s = stepsOf(id)
assert.equal(s[0].status, 'done')
assert.equal(s[1].status, 'done')
assert.equal(s[2].status, 'pending', 'the step after a wait must not run in the same tick')
assert.equal(s[2].due_at.getTime(), later(300_000).getTime(), 'the wait is the NEXT step due_at')
assert.equal(run(id).status, 'running')
assert.equal(run(id).claimed_by, null, 'a run left in flight gives its lease back')
// Tick two, still inside the wait: nothing moves.
await runner.tick(later(120_000))
assert.equal(stepsOf(id)[2].status, 'pending')
assert.equal(run(id).status, 'running')
// Tick three, past it: the last step runs and the run completes.
await runner.tick(later(301_000))
assert.equal(stepsOf(id)[2].status, 'done')
assert.equal(run(id).status, 'completed')
assert.equal(run(id).health, 'ok')
assert.ok(kinds(id).includes('phase.completed'))
assert.ok(kinds(id).includes('run.status'))
})
test('a run passes through `ending` on its way to completed', async () => {
register([scriptedAction('test.noop')])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.noop')] }])
await runner.tick(T0)
const transitions = store.log.filter((l) => l.runId === id && l.kind === 'run.status').map((l) => l.detail.to)
assert.deepEqual(transitions, ['starting', 'running', 'ending', 'completed'])
})
test('phases run in order and the next one is materialised on entry', async () => {
register([scriptedAction('test.noop')])
const id = seedRun([
{ key: 'opening', label: 'Opening', steps: [step('test.noop')] },
{ key: 'closing', label: 'Closing', steps: [step('test.noop'), step('test.noop')] },
])
await runner.tick(T0)
assert.equal(run(id).status, 'completed')
assert.deepEqual(stepsOf(id).map((s) => s.phase), ['opening', 'closing', 'closing'])
assert.ok(stepsOf(id).every((s) => s.status === 'done'))
})
test('a wait as the last step of a phase holds the NEXT phase, rather than meaning nothing', async () => {
register([scriptedAction('test.noop'), scriptedAction('test.wait')])
scripted['test.wait'] = { calls: [], answer: { ok: true, holdFor: 300 } }
const id = seedRun([
{ key: 'opening', label: 'Opening', steps: [step('test.noop'), step('test.wait')] },
{ key: 'closing', label: 'Closing', steps: [step('test.noop')] },
])
await runner.tick(T0)
const closing = stepsOf(id).filter((x) => x.phase === 'closing')
assert.equal(closing.length, 1, 'the next phase is materialised')
assert.equal(closing[0].status, 'pending')
assert.equal(
closing[0].due_at.getTime(),
later(300_000).getTime(),
'the hold crosses the phase boundary; dropping it would start the next phase at once',
)
assert.equal(run(id).status, 'running')
await runner.tick(later(301_000))
assert.equal(run(id).status, 'completed')
})
test('a GM cue parks: the step stays running with no lease, and the reclaim leaves it alone', async () => {
register([scriptedAction('test.cue'), scriptedAction('test.after')])
scripted['test.cue'] = { calls: [], answer: { ok: true, await: 'human' } }
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.cue'), step('test.after')] }])
await runner.tick(T0)
const cue = stepsOf(id)[0]
assert.equal(cue.status, 'running')
assert.equal(cue.claim_expires_at, null, 'a parked step carries no lease')
assert.equal(stepsOf(id)[1].status, 'pending', 'nothing after a cue proceeds')
assert.ok(kinds(id).includes('step.parked'))
// A week later the reclaim has still not touched it, and the cue has been
// dispatched exactly once. This is the whole point of a NULL lease.
await runner.tick(later(7 * 24 * 60 * 60 * 1000))
assert.equal(stepsOf(id)[0].status, 'running')
assert.equal(stepsOf(id)[0].attempts, 1)
assert.equal(scripted['test.cue'].calls.length, 1)
assert.equal(run(id).status, 'running')
})
test('a reclaim returns a stale step without resetting attempts', async () => {
register([scriptedAction('test.slow')])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.slow')] }])
// Simulate a process that claimed the step and died: running, lease in the past.
await runner.tick(T0)
const s = stepsOf(id)[0]
Object.assign(store.steps.get(s.id), { status: 'running', attempts: 2, claim_expires_at: later(-1000) })
await stepsDb.reclaimStale(T0, runner.MAX_ATTEMPTS)
assert.equal(store.steps.get(s.id).status, 'pending')
assert.equal(store.steps.get(s.id).attempts, 2, 'Engagement Phase 14: a reclaim must never reset attempts')
})
test('a step whose attempts are spent leaves `running` as failed rather than being handed back', async () => {
register([scriptedAction('test.slow')])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.slow')] }])
await runner.tick(T0)
const s = stepsOf(id)[0]
Object.assign(store.steps.get(s.id), { status: 'running', attempts: runner.MAX_ATTEMPTS, claim_expires_at: later(-1000) })
const { failed, reclaimed } = await stepsDb.reclaimStale(T0, runner.MAX_ATTEMPTS)
assert.equal(failed, 1)
assert.equal(reclaimed, 0, 'a row that gave up must not also be reclaimed, or it retries forever')
assert.equal(store.steps.get(s.id).status, 'failed')
})
test('a transient failure retries on a flat backoff with the same idempotency key, then applies on_failure', async () => {
register([scriptedAction('test.flaky')])
scripted['test.flaky'] = { calls: [], answer: { ok: false, retry: true, error: 'relay is down' } }
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.flaky', {}, 'skip')] }])
const key = () => stepsOf(id)[0].idempotency_key
await runner.tick(T0)
const firstKey = key()
assert.equal(stepsOf(id)[0].status, 'pending')
assert.equal(stepsOf(id)[0].attempts, 1)
assert.equal(stepsOf(id)[0].due_at.getTime(), later(runner.RETRY_MS).getTime())
assert.equal(run(id).health, 'degraded', 'degraded from the first retry, not from the eventual failure')
await runner.tick(later(runner.RETRY_MS))
assert.equal(stepsOf(id)[0].attempts, 2)
await runner.tick(later(2 * runner.RETRY_MS))
assert.equal(stepsOf(id)[0].attempts, runner.MAX_ATTEMPTS)
assert.equal(stepsOf(id)[0].status, 'failed', 'all three dispositions write the step failed')
assert.equal(run(id).status, 'completed', 'on_failure: skip lets the run finish')
assert.equal(run(id).health, 'degraded')
assert.equal(key(), firstKey, 'the idempotency key does not vary by attempt')
assert.equal(new Set(scripted['test.flaky'].calls.map((c) => c.idempotencyKey)).size, 1)
})
test('on_failure: pause stops the run and the tick never picks it up again', async () => {
register([scriptedAction('test.bad'), scriptedAction('test.after')])
scripted['test.bad'] = { calls: [], answer: { ok: false, retry: false, error: 'the world is half changed' } }
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.bad', {}, 'pause'), step('test.after')] }])
await runner.tick(T0)
assert.equal(run(id).status, 'paused')
assert.equal(stepsOf(id)[0].status, 'failed')
assert.equal(stepsOf(id)[1].status, 'pending', 'a paused run leaves its remaining steps alone')
await runner.tick(later(60_000))
assert.equal(run(id).status, 'paused', 'only Phase 3 resume moves a paused run')
assert.equal(scripted['test.after']?.calls?.length ?? 0, 0)
})
test('on_failure: abort_run fails the run and cancels what has not started', async () => {
register([scriptedAction('test.bad'), scriptedAction('test.after')])
scripted['test.bad'] = { calls: [], answer: { ok: false, retry: false, error: 'no' } }
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.bad', {}, 'abort_run'), step('test.after')] }])
await runner.tick(T0)
assert.equal(run(id).status, 'failed')
assert.equal(stepsOf(id)[0].status, 'failed')
assert.equal(stepsOf(id)[1].status, 'cancelled')
})
test('a step naming an unregistered action fails terminal with the module named, and degrades the run', async () => {
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('gone.verb', {}, 'skip')] }])
await runner.tick(T0)
const s = stepsOf(id)[0]
assert.equal(s.status, 'failed', 'never a silent skip (§L)')
assert.equal(s.attempts, 1, 'a dormant action is terminal, so it is not retried')
assert.match(s.last_error, /gone\.verb/)
assert.equal(run(id).health, 'degraded')
})
test('a held concurrency key holds the run at scheduled, and logs the reason once', async () => {
register([scriptedAction('test.noop'), scriptedAction('test.cue')])
scripted['test.cue'] = { calls: [], answer: { ok: true, await: 'human' } }
// The holder is parked on a cue, which is what keeps it genuinely in flight. A
// holder with no steps would complete itself on this same tick — correct
// behaviour, and a fixture that proved nothing.
const holder = seedRun([{ key: 'main', label: 'Main', steps: [step('test.cue')] }], {
concurrencyKey: 'invasion:Yew',
})
const waiting = seedRun([{ key: 'main', label: 'Main', steps: [step('test.noop')] }], { concurrencyKey: 'invasion:Yew' })
await runner.tick(T0)
assert.equal(run(waiting).status, 'scheduled')
assert.match(run(waiting).last_error, new RegExp(`run ${holder}`))
assert.equal(kinds(waiting).filter((k) => k === 'run.blocked').length, 1)
// Still held, and still one line: a line per tick would bury the one that matters.
await runner.tick(later(15_000))
assert.equal(kinds(waiting).filter((k) => k === 'run.blocked').length, 1)
// The holder finishes, and the next tick starts the run that was waiting.
store.runs.get(holder).status = 'completed'
store.runs.get(holder).claim_expires_at = null
await runner.tick(later(30_000))
assert.equal(run(waiting).status, 'completed')
})
test('an occurrence past its own grace window is missed, never a late silent start', async () => {
register([scriptedAction('test.noop')])
const late = seedRun([{ key: 'main', label: 'Main', steps: [step('test.noop')] }], { graceSeconds: 600 })
const inside = seedRun([{ key: 'main', label: 'Main', steps: [step('test.noop')] }], { graceSeconds: 3600 })
// Both are due; the process has been down for half an hour.
await runner.tick(later(30 * 60 * 1000))
assert.equal(run(late).status, 'missed')
assert.equal(stepsOf(late)[0].status, 'cancelled')
assert.equal(run(inside).status, 'completed', 'inside its window it starts late and says so')
})
test('a run in `ending` when the process died is completed by the next tick', async () => {
const id = seedRun([{ key: 'main', label: 'Main', steps: [] }], { status: 'ending' })
store.runs.get(id).current_phase = 'main'
await runner.tick(T0)
assert.equal(run(id).status, 'completed')
})
test('a live lease is not re-enterable, not even by the process that took it', async () => {
register([scriptedAction('test.cue')])
scripted['test.cue'] = { calls: [], answer: { ok: true, await: 'human' } }
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.cue')] }])
await runner.tick(T0)
// Put a live lease back on the run, as an overrunning tick would have.
Object.assign(store.runs.get(id), { claimed_by: runner.OWNER, claim_expires_at: later(60_000) })
const taken = await runsDb.claimTick(id, runner.OWNER, later(120_000), T0)
assert.equal(taken, false, 'the CAS is what protects a tick that overran into the next one')
})
// ── §F: no shape a failure can take reads as success ───────────────────────
test('classify: every failure shape is a failure', () => {
assert.equal(classify(undefined, 'a').outcome, 'retry')
assert.equal(classify(null, 'a').outcome, 'retry')
assert.equal(classify('ok', 'a').outcome, 'retry')
assert.equal(classify(['ok'], 'a').outcome, 'retry')
assert.equal(classify({}, 'a').outcome, 'retry', 'a missing ok is not a success')
assert.equal(classify({ ok: 'yes' }, 'a').outcome, 'retry', 'ok must be true, not truthy')
assert.equal(classify({ ok: false }, 'a').outcome, 'retry')
assert.equal(classify({ ok: false, retry: false }, 'a').outcome, 'terminal')
assert.equal(classify({ __timedOut: true, error: 'slow' }, 'a').outcome, 'retry')
})
test('classify: the two success shapes that are not "finished"', () => {
assert.equal(classify({ ok: true }, 'a').outcome, 'done')
assert.equal(classify({ ok: true }, 'a').holdSeconds, 0)
assert.equal(classify({ ok: true, await: 'human' }, 'a').outcome, 'parked')
assert.equal(classify({ ok: true, holdFor: 90 }, 'a').holdSeconds, 90)
assert.equal(classify({ ok: true, holdFor: '90' }, 'a').holdSeconds, 90)
assert.equal(classify({ ok: true, holdFor: -1 }, 'a').outcome, 'terminal', 'a bad holdFor is not a silent zero')
assert.equal(classify({ ok: true, holdFor: 'soon' }, 'a').outcome, 'terminal')
assert.ok(classify({ ok: true, holdFor: 1e12 }, 'a').holdSeconds <= 7 * 24 * 60 * 60, 'holdFor is bounded')
})
test('an action that throws is a transient failure, not a crashed tick', async () => {
register([
scriptedAction('test.thrower', {
perform: async () => {
throw new Error('boom')
},
}),
])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.thrower', {}, 'skip')] }])
await runner.tick(T0)
assert.equal(stepsOf(id)[0].status, 'pending')
assert.match(stepsOf(id)[0].last_error, /boom/)
assert.equal(run(id).status, 'running', 'one bad action does not stop the deployment')
})
test('an action that never answers is cut off at its declared budget', async () => {
register([
scriptedAction('test.hang', { budgetMs: 30, perform: () => new Promise(() => {}) }),
])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.hang', {}, 'skip')] }])
await runner.tick(T0)
assert.equal(stepsOf(id)[0].status, 'pending', 'a timeout is transient')
assert.match(stepsOf(id)[0].last_error, /budget/)
})
test('the dispatch envelope carries what §F says it carries', async () => {
register([scriptedAction('test.echo')])
const id = seedRun([{ key: 'main', label: 'Main', steps: [step('test.echo', {})] }])
store.runs.get(id).scope = 'atlantic'
await runner.tick(T0)
const [envelope] = scripted['test.echo'].calls
assert.equal(envelope.runId, id)
assert.equal(envelope.scope, 'atlantic')
assert.equal(envelope.verify, false)
assert.equal(typeof envelope.idempotencyKey, 'string')
assert.equal(envelope.idempotencyKey.length, 40)
assert.deepEqual(Object.keys(envelope).sort(), ['actor', 'idempotencyKey', 'params', 'runId', 'scope', 'stepId', 'verify'])
})

View File

@@ -0,0 +1,508 @@
// ── The runner's raw SQL, against a real MariaDB ───────────────────────────
//
// EVENTS_PLAN.md Phase 2. `eventRunner.test.js` stubs the three tables and
// exercises everything the runner DECIDES. It cannot prove the statements whose
// whole correctness is a server contract, and on this codebase that gap has
// already cost something once: engagement's cooldown claim was green against its
// stub and always allowed the send against a real server, because the connector
// defaults `foundRows: true` and a no-op UPDATE reports 1 rather than 0.
//
// So the five statements that decide who owns what run here, for real:
//
// • **`claimStart`** — the CAS `scheduled -> starting`. "Exactly one winner" is
// `affectedRows = 1` for one caller and 0 for every other, and that is a
// property of the SERVER's answer, not of the SQL's shape.
// • **`claimTick`** — the same, for a run already in flight, and with no
// owner-matches escape clause. A live lease must refuse its own holder, or
// one `setInterval` that overran advances one run twice.
// • **`transition`** — a guarded status move. The guard is the whole thing: a
// run cancelled between the read and the write must not be transitioned.
// • **`reclaimStale`** on steps — two statements in a fixed ORDER, give-up
// before hand-back. Reversing them makes `MAX_ATTEMPTS` unreachable and the
// row cycles forever (Engagement Phase 14's defect), and **neither statement
// may touch `attempts`**.
// • **`holdNext`** — `UPDATE ... ORDER BY seq LIMIT 1` with a guard, which is
// both a MariaDB-specific syntax and a correctness claim: it must move the
// next PENDING step and only ever push a due date later.
//
// Plus the two unique indexes that are load-bearing rather than tidy:
// `uq_evrun_occurrence` (which, not the claim, is what stops two runs of one
// occurrence existing) and `uq_evstep_slot` (which is what makes re-materialising
// a phase a no-op).
//
// **It SKIPS when there is no database**, deliberately: CI runs the suite with
// the pool pointed at a dead port, and a file that failed there would make every
// PR red for a reason unrelated to itself. Run it against this machine's
// container with:
//
// DB_HOST=127.0.0.1 DB_PORT=3306 DB_USER=... DB_PASSWORD=... \
// node --test test/eventRunnerSql.test.js
//
// It creates its tables in a throwaway database named after the process and
// drops it again, so it can never touch a real schema.
const { test, before, after, beforeEach } = require('node:test')
const assert = require('node:assert/strict')
const mariadb = require('mariadb')
// Trimmed to the columns these statements read or write. The ENUMs are verbatim,
// because "is `missed` a legal value" is one of the things being proved.
const SCHEMA = `
CREATE TABLE event_definitions (
id INT AUTO_INCREMENT PRIMARY KEY,
grace_seconds INT NOT NULL DEFAULT 900
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
CREATE TABLE event_runs (
id BIGINT AUTO_INCREMENT PRIMARY KEY,
definition_id INT NOT NULL,
version_id INT NOT NULL,
scope VARCHAR(190) NOT NULL DEFAULT '',
status ENUM('scheduled','starting','running','paused','ending',
'completed','cancelled','failed','missed')
NOT NULL DEFAULT 'scheduled',
health ENUM('ok','degraded','stalled') NOT NULL DEFAULT 'ok',
current_phase VARCHAR(64) NULL,
scheduled_for DATETIME NOT NULL,
concurrency_key VARCHAR(190) NULL,
started_at DATETIME NULL,
ended_at DATETIME NULL,
claimed_by VARCHAR(64) NULL,
claim_expires_at DATETIME NULL,
last_error VARCHAR(500) NULL,
created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
UNIQUE KEY uq_evrun_occurrence (definition_id, scope, scheduled_for),
INDEX idx_evrun_due (status, scheduled_for),
INDEX idx_evrun_concurrency (concurrency_key, status)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
CREATE TABLE event_run_steps (
id BIGINT AUTO_INCREMENT PRIMARY KEY,
run_id BIGINT NOT NULL,
phase VARCHAR(64) NOT NULL,
seq INT NOT NULL,
action_id VARCHAR(96) NOT NULL,
params JSON NULL,
action_version INT NOT NULL DEFAULT 1,
status ENUM('pending','running','done','failed','skipped','refused','cancelled')
NOT NULL DEFAULT 'pending',
due_at DATETIME NULL,
attempts INT NOT NULL DEFAULT 0,
on_failure VARCHAR(32) NOT NULL DEFAULT 'skip',
idempotency_key CHAR(40) NOT NULL,
claimed_by VARCHAR(64) NULL,
claim_expires_at DATETIME NULL,
last_error VARCHAR(500) NULL,
started_at DATETIME NULL,
finished_at DATETIME NULL,
created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
updated_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
UNIQUE KEY uq_evstep_slot (run_id, phase, seq),
INDEX idx_evstep_due (status, due_at)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
`
// The statements under test, verbatim from `eventRuns.db.js` and
// `eventRunSteps.db.js`. Duplicated rather than required, because requiring the
// modules would drag in `utils/db`'s pool, which the harness has already pointed
// at a dead port. The pool below leaves `foundRows` at the connector's default,
// exactly as `utils/db.js` does — pinning it here would make this file agree with
// the code by construction and prove nothing about the pool the server runs.
const CLAIM_START = `
UPDATE event_runs
SET status = 'starting', claimed_by = ?, claim_expires_at = ?,
started_at = COALESCE(started_at, NOW())
WHERE id = ? AND status = 'scheduled'`
const CLAIM_TICK = `
UPDATE event_runs
SET claimed_by = ?, claim_expires_at = ?
WHERE id = ?
AND status IN ('starting','running','ending')
AND (claim_expires_at IS NULL OR claim_expires_at < ?)`
const TRANSITION = `
UPDATE event_runs SET status = ?, current_phase = ?
WHERE id = ? AND status IN (?)`
const CLAIM_STEP = `
UPDATE event_run_steps
SET status = 'running', attempts = attempts + 1, claimed_by = ?, claim_expires_at = ?,
started_at = COALESCE(started_at, NOW())
WHERE id = ? AND status = 'pending' AND (due_at IS NULL OR due_at <= ?)`
const STEP_GIVE_UP = `
UPDATE event_run_steps
SET status = 'failed', last_error = 'gave up after repeated interruptions',
finished_at = NOW(), claimed_by = NULL, claim_expires_at = NULL
WHERE status = 'running'
AND claim_expires_at IS NOT NULL AND claim_expires_at < ?
AND attempts >= ?`
const STEP_RECLAIM = `
UPDATE event_run_steps
SET status = 'pending', claimed_by = NULL, claim_expires_at = NULL
WHERE status = 'running' AND claim_expires_at IS NOT NULL AND claim_expires_at < ?`
const HOLD_NEXT = `
UPDATE event_run_steps
SET due_at = ?
WHERE run_id = ? AND phase = ? AND seq > ? AND status = 'pending'
AND (due_at IS NULL OR due_at < ?)
ORDER BY seq LIMIT 1`
const MATERIALISE_RUN = `
INSERT IGNORE INTO event_runs (definition_id, version_id, scope, scheduled_for, concurrency_key)
VALUES (?, ?, ?, ?, ?)`
const MATERIALISE_STEP = `
INSERT IGNORE INTO event_run_steps (run_id, phase, seq, action_id, idempotency_key)
VALUES (?, ?, ?, ?, ?)`
const FIND_MISSED = `
SELECT r.id FROM event_runs r
JOIN event_definitions d ON d.id = r.definition_id
WHERE r.status = 'scheduled'
AND r.scheduled_for + INTERVAL d.grace_seconds SECOND < ?`
const DB = `rg_events_test_${process.pid}`
let pool = null
let available = false
const poolOpts = () => ({
host: process.env.DB_HOST || '127.0.0.1',
port: Number(process.env.DB_PORT) || 3306,
user: process.env.DB_USER || 'root',
password: process.env.DB_PASSWORD || '',
})
before(async () => {
const admin = mariadb.createPool({
...poolOpts(),
connectionLimit: 1,
connectTimeout: 2000,
initializationTimeout: 2000,
multipleStatements: true,
})
try {
await admin.query(`CREATE DATABASE ${DB}`)
available = true
} catch {
available = false
} finally {
await admin.end().catch(() => {})
}
if (!available) return
pool = mariadb.createPool({
...poolOpts(),
database: DB,
connectionLimit: 3,
multipleStatements: true,
bigIntAsNumber: true,
insertIdAsNumber: true,
})
await pool.query(SCHEMA)
})
after(async () => {
if (pool) {
await pool.query(`DROP DATABASE IF EXISTS ${DB}`).catch(() => {})
await pool.end().catch(() => {})
}
})
// Checked INSIDE each test, never as a `{ skip }` option: the option is evaluated
// when the file is read, which is before `before()` has had a chance to find out
// whether there is a database. Every test skipped unconditionally is what that
// mistake looks like, and it looks exactly like a passing suite.
const SKIP = 'no database reachable - set DB_HOST/DB_PORT/DB_USER/DB_PASSWORD to run'
const needDb = (t) => {
if (available) return false
t.skip(SKIP)
return true
}
const T0 = new Date('2026-09-02T12:00:00Z')
const later = (ms) => new Date(T0.getTime() + ms)
const rows = (r) => Number(r.affectedRows)
beforeEach(async () => {
if (!available) return
await pool.query('DELETE FROM event_run_steps')
await pool.query('DELETE FROM event_runs')
await pool.query('DELETE FROM event_definitions')
})
async function seedRun(over = {}) {
const def = await pool.query('INSERT INTO event_definitions (grace_seconds) VALUES (?)', [
over.graceSeconds ?? 900,
])
const r = await pool.query(
`INSERT INTO event_runs (definition_id, version_id, scope, status, scheduled_for, concurrency_key,
claimed_by, claim_expires_at)
VALUES (?, 1, ?, ?, ?, ?, ?, ?)`,
[
def.insertId,
over.scope ?? '',
over.status ?? 'scheduled',
over.scheduledFor ?? T0,
over.concurrencyKey ?? null,
over.claimedBy ?? null,
over.claimExpiresAt ?? null,
],
)
return { runId: r.insertId, definitionId: def.insertId }
}
const seedStep = async (runId, over = {}) =>
(
await pool.query(
`INSERT INTO event_run_steps (run_id, phase, seq, action_id, status, due_at, attempts,
claimed_by, claim_expires_at, idempotency_key)
VALUES (?, ?, ?, 'test.noop', ?, ?, ?, ?, ?, ?)`,
[
runId,
over.phase ?? 'main',
over.seq ?? 0,
over.status ?? 'pending',
over.dueAt ?? null,
over.attempts ?? 0,
over.claimedBy ?? null,
over.claimExpiresAt ?? null,
over.key ?? 'k'.repeat(40),
],
)
).insertId
const stepById = async (id) => (await pool.query('SELECT * FROM event_run_steps WHERE id = ?', [id]))[0]
const runById = async (id) => (await pool.query('SELECT * FROM event_runs WHERE id = ?', [id]))[0]
// ── claimStart: exactly one winner ─────────────────────────────────────────
test('claimStart: the first caller wins and every other gets zero', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun()
const first = await pool.query(CLAIM_START, ['host:1', later(60_000), runId])
const second = await pool.query(CLAIM_START, ['host:2', later(60_000), runId])
assert.equal(rows(first), 1, 'the winner is told 1')
assert.equal(rows(second), 0, 'the loser is told 0, not 1 with foundRows')
assert.equal((await runById(runId)).claimed_by, 'host:1')
assert.equal((await runById(runId)).status, 'starting')
})
test('claimStart: started_at is stamped once and never moved', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun()
await pool.query(CLAIM_START, ['host:1', later(60_000), runId])
const first = (await runById(runId)).started_at
await pool.query("UPDATE event_runs SET status = 'scheduled' WHERE id = ?", [runId])
await pool.query(CLAIM_START, ['host:2', later(60_000), runId])
assert.deepEqual((await runById(runId)).started_at, first, 'COALESCE keeps the original instant')
})
// ── claimTick: a live lease refuses even its own holder ────────────────────
test('claimTick: a live lease is not re-enterable by the process that took it', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running', claimedBy: 'host:1', claimExpiresAt: later(60_000) })
const again = await pool.query(CLAIM_TICK, ['host:1', later(120_000), runId, T0])
assert.equal(rows(again), 0, 'a tick that overran must not advance its own run twice')
})
test('claimTick: an expired lease is takeable, by anyone', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running', claimedBy: 'host:1', claimExpiresAt: later(-60_000) })
const taken = await pool.query(CLAIM_TICK, ['host:2', later(60_000), runId, T0])
assert.equal(rows(taken), 1)
assert.equal((await runById(runId)).claimed_by, 'host:2')
})
test('claimTick: a paused run is never claimable', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'paused' })
assert.equal(rows(await pool.query(CLAIM_TICK, ['host:1', later(60_000), runId, T0])), 0)
})
// ── transition: the guard is the whole point ───────────────────────────────
test('transition: a run cancelled underneath the tick is not transitioned', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
await pool.query("UPDATE event_runs SET status = 'cancelled' WHERE id = ?", [runId])
const moved = await pool.query(TRANSITION, ['completed', 'main', runId, 'running'])
assert.equal(rows(moved), 0)
assert.equal((await runById(runId)).status, 'cancelled')
})
test('transition: running -> running is a guarded write, not a no-op', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
// This is how the runner advances `current_phase`, and `foundRows` is exactly
// what makes it report 1 despite `status` not changing — which is the answer
// the caller needs, because what it is checking is that the run is STILL
// running, not that the status moved.
const moved = await pool.query(TRANSITION, ['running', 'closing', runId, 'running'])
assert.equal(rows(moved), 1)
assert.equal((await runById(runId)).current_phase, 'closing')
})
// ── The step claim ─────────────────────────────────────────────────────────
test('the step claim: one winner, and attempts is incremented by the claim alone', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
const stepId = await seedStep(runId)
assert.equal(rows(await pool.query(CLAIM_STEP, ['host:1', later(60_000), stepId, T0])), 1)
assert.equal(rows(await pool.query(CLAIM_STEP, ['host:2', later(60_000), stepId, T0])), 0)
assert.equal((await stepById(stepId)).attempts, 1, 'one claim, one attempt')
})
test('the step claim: a step held behind a core.wait is not due', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
const stepId = await seedStep(runId, { dueAt: later(300_000) })
assert.equal(rows(await pool.query(CLAIM_STEP, ['host:1', later(60_000), stepId, T0])), 0)
assert.equal(rows(await pool.query(CLAIM_STEP, ['host:1', later(360_000), stepId, later(301_000)])), 1)
})
// ── The reclaim: order, and what it must not touch ─────────────────────────
test('the reclaim hands a stale step back WITHOUT resetting attempts', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
const stepId = await seedStep(runId, { status: 'running', attempts: 2, claimExpiresAt: later(-1000) })
await pool.query(STEP_GIVE_UP, [T0, 3])
await pool.query(STEP_RECLAIM, [T0])
const step = await stepById(stepId)
assert.equal(step.status, 'pending')
assert.equal(step.attempts, 2, 'Engagement Phase 14: a reclaim that reset this made MAX_ATTEMPTS unreachable')
})
test('the reclaim gives up FIRST, so a spent step leaves running as failed', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
const stepId = await seedStep(runId, { status: 'running', attempts: 3, claimExpiresAt: later(-1000) })
const gaveUp = await pool.query(STEP_GIVE_UP, [T0, 3])
const reclaimed = await pool.query(STEP_RECLAIM, [T0])
assert.equal(rows(gaveUp), 1)
assert.equal(rows(reclaimed), 0, 'reversing these two makes the row retry forever')
assert.equal((await stepById(stepId)).status, 'failed')
assert.equal((await stepById(stepId)).attempts, 3)
})
test('the reclaim leaves a PARKED step alone, however long it waits', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
const stepId = await seedStep(runId, { status: 'running', attempts: 1, claimExpiresAt: null })
await pool.query(STEP_GIVE_UP, [later(365 * 24 * 3600 * 1000), 3])
await pool.query(STEP_RECLAIM, [later(365 * 24 * 3600 * 1000)])
const step = await stepById(stepId)
assert.equal(step.status, 'running', 'a NULL lease is a parked cue, not staleness')
assert.equal(step.attempts, 1)
})
// ── holdNext ───────────────────────────────────────────────────────────────
test('holdNext moves the next PENDING step of the phase, and only one', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
await seedStep(runId, { seq: 0, status: 'done' })
const second = await seedStep(runId, { seq: 1 })
const third = await seedStep(runId, { seq: 2 })
const moved = await pool.query(HOLD_NEXT, [later(300_000), runId, 'main', 0, later(300_000)])
assert.equal(rows(moved), 1)
assert.deepEqual((await stepById(second)).due_at, later(300_000))
assert.equal((await stepById(third)).due_at, null, 'a wait holds the next step, not the rest of the phase')
})
test('holdNext never pulls a due date earlier, so a re-dispatch cannot double the wait', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
await seedStep(runId, { seq: 0, status: 'done' })
const second = await seedStep(runId, { seq: 1, dueAt: later(600_000) })
const moved = await pool.query(HOLD_NEXT, [later(300_000), runId, 'main', 0, later(300_000)])
assert.equal(rows(moved), 0)
assert.deepEqual((await stepById(second)).due_at, later(600_000))
})
test('holdNext skips a step that is already running', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
await seedStep(runId, { seq: 0, status: 'done' })
await seedStep(runId, { seq: 1, status: 'running' })
const third = await seedStep(runId, { seq: 2 })
await pool.query(HOLD_NEXT, [later(300_000), runId, 'main', 0, later(300_000)])
assert.deepEqual((await stepById(third)).due_at, later(300_000))
})
// ── The two unique indexes that carry the weight ───────────────────────────
test('uq_evrun_occurrence, not the claim, is what stops two runs of one occurrence', async (t) => {
if (needDb(t)) return
const def = await pool.query('INSERT INTO event_definitions (grace_seconds) VALUES (900)')
const a = await pool.query(MATERIALISE_RUN, [def.insertId, 1, '', T0, null])
const b = await pool.query(MATERIALISE_RUN, [def.insertId, 1, '', T0, null])
assert.equal(rows(a), 1)
assert.equal(rows(b), 0, 'INSERT IGNORE answers honestly rather than raising a 1062')
const all = await pool.query('SELECT COUNT(*) AS n FROM event_runs')
assert.equal(Number(all[0].n), 1)
})
test("scope '' rather than NULL is what makes that index work at all", async (t) => {
if (needDb(t)) return
const def = await pool.query('INSERT INTO event_definitions (grace_seconds) VALUES (900)')
// The empty-string case collides, as it must. A NULL scope would NOT: multiple
// NULLs do not collide in MariaDB, which would silently permit two runs of one
// occurrence — the reason the column is NOT NULL DEFAULT ''.
await pool.query(MATERIALISE_RUN, [def.insertId, 1, '', T0, null])
assert.equal(rows(await pool.query(MATERIALISE_RUN, [def.insertId, 1, '', T0, null])), 0)
// Two different scopes are two different occurrences, which is what lets a
// worldwide event fan out across servers without colliding with itself.
assert.equal(rows(await pool.query(MATERIALISE_RUN, [def.insertId, 1, 'atlantic', T0, null])), 1)
})
test('uq_evstep_slot makes re-materialising a phase a no-op', async (t) => {
if (needDb(t)) return
const { runId } = await seedRun({ status: 'running' })
assert.equal(rows(await pool.query(MATERIALISE_STEP, [runId, 'main', 0, 'test.noop', 'a'.repeat(40)])), 1)
assert.equal(rows(await pool.query(MATERIALISE_STEP, [runId, 'main', 0, 'test.noop', 'b'.repeat(40)])), 0)
const [step] = await pool.query('SELECT idempotency_key FROM event_run_steps WHERE run_id = ?', [runId])
assert.equal(step.idempotency_key, 'a'.repeat(40), 'a re-materialise cannot overwrite a key a dispatch already sent')
})
// ── The grace window is per definition ─────────────────────────────────────
test('findMissed compares against each definitions own grace window', async (t) => {
if (needDb(t)) return
const tight = await seedRun({ graceSeconds: 600 })
const generous = await seedRun({ graceSeconds: 3600, scope: 'b' })
const missed = (await pool.query(FIND_MISSED, [later(30 * 60 * 1000)])).map((r) => Number(r.id))
assert.deepEqual(missed, [tight.runId])
assert.ok(!missed.includes(generous.runId), 'inside its own window a run starts late rather than being missed')
})