feat(ql3): report legacy shadow startup differences

This commit is contained in:
whyour
2026-08-18 09:58:16 +08:00
parent 6cb9a075d7
commit d38f624262
9 changed files with 722 additions and 85 deletions
@@ -6,7 +6,12 @@ import {
} from '../../domain/deploymentProfile';
import { configuredLegacyShadowOrigins } from '../../compatibility/legacyExecutionBridge';
import { createLegacyLogArtifactId } from '../../compatibility/legacyTaskRevision';
import type { LegacyShadowStartupSummary } from '../../application/legacyShadowStartupReconciler';
import {
LEGACY_SHADOW_STARTUP_OUTCOMES,
type LegacyShadowStartupOriginSummary,
type LegacyShadowStartupOutcomeCounts,
type LegacyShadowStartupSummary,
} from '../../application/legacyShadowStartupReconciler';
export interface LegacyShadowStartupBudget {
pageSize: number;
@@ -18,6 +23,60 @@ export interface LegacyShadowStartupRequest extends LegacyShadowStartupBudget {
profile: 'edge' | 'standalone';
}
export const LEGACY_SHADOW_STARTUP_DIFFERENCE_REPORT_SCHEMA =
'qinglong/legacy-shadow-startup-difference-report@v1';
export const LEGACY_SHADOW_STARTUP_METRIC_BATCH_SCHEMA =
'qinglong/legacy-shadow-startup-metric-batch@v1';
export type LegacyShadowStartupAssessment =
| 'converged'
| 'waiting_external_callback'
| 'incomplete'
| 'attention_required';
export interface LegacyShadowStartupDifferenceReport {
schemaVersion: 1;
schema: typeof LEGACY_SHADOW_STARTUP_DIFFERENCE_REPORT_SCHEMA;
profile: 'edge' | 'standalone';
assessment: LegacyShadowStartupAssessment;
configuredOriginCount: number;
budget: {
pageSize: number;
maxPages: number;
maxCandidates: number;
};
coverage: {
pages: number;
scanned: number;
stopReason: LegacyShadowStartupSummary['stopReason'];
remaining: boolean;
resumeAvailable: boolean;
};
outcomes: LegacyShadowStartupOutcomeCounts;
byOrigin: readonly LegacyShadowStartupOriginSummary[];
}
export interface LegacyShadowStartupMetricBatch {
schemaVersion: 1;
schema: typeof LEGACY_SHADOW_STARTUP_METRIC_BATCH_SCHEMA;
dimensions: {
profile: 'edge' | 'standalone';
assessment: LegacyShadowStartupAssessment;
stopReason: LegacyShadowStartupSummary['stopReason'];
};
values: LegacyShadowStartupOutcomeCounts & {
configuredOrigins: number;
pageSize: number;
maxPages: number;
maxCandidates: number;
pages: number;
scanned: number;
remaining: 0 | 1;
resumeAvailable: 0 | 1;
};
byOrigin: readonly LegacyShadowStartupOriginSummary[];
}
export type LegacyShadowStartupAudit =
| {
state: 'disabled';
@@ -30,9 +89,11 @@ export type LegacyShadowStartupAudit =
state: 'reconciled' | 'incomplete';
profile: 'edge' | 'standalone';
origins: number;
summary: Omit<LegacyShadowStartupSummary, 'nextCursor'> & {
summary: Omit<LegacyShadowStartupSummary, 'nextCursor' | 'byOrigin'> & {
resumeAvailable: boolean;
};
report: LegacyShadowStartupDifferenceReport;
metrics: LegacyShadowStartupMetricBatch;
}
| {
state: 'failed';
@@ -46,6 +107,7 @@ export interface BootstrapLegacyShadowStartupOptions {
request: LegacyShadowStartupRequest,
) => Promise<LegacyShadowStartupSummary>;
audit?: (record: LegacyShadowStartupAudit) => void | Promise<void>;
collect?: (metrics: LegacyShadowStartupMetricBatch) => void | Promise<void>;
}
const BUDGETS: Readonly<
@@ -54,17 +116,195 @@ const BUDGETS: Readonly<
edge: { pageSize: 8, maxPages: 1 },
standalone: { pageSize: 32, maxPages: 4 },
};
const MAX_CONFIGURED_LEGACY_SHADOW_ORIGINS = 7;
function auditSummary(summary: LegacyShadowStartupSummary): Omit<
LegacyShadowStartupSummary,
'nextCursor'
'nextCursor' | 'byOrigin'
> & {
resumeAvailable: boolean;
} {
const { nextCursor, ...bounded } = summary;
const { nextCursor, byOrigin: _byOrigin, ...bounded } = summary;
return { ...bounded, resumeAvailable: nextCursor !== undefined };
}
function outcomeCounts(
source: LegacyShadowStartupOutcomeCounts,
): LegacyShadowStartupOutcomeCounts {
return Object.fromEntries(
LEGACY_SHADOW_STARTUP_OUTCOMES.map((outcome) => [outcome, source[outcome]]),
) as LegacyShadowStartupOutcomeCounts;
}
function outcomeTotal(source: LegacyShadowStartupOutcomeCounts): number {
return LEGACY_SHADOW_STARTUP_OUTCOMES.reduce(
(total, outcome) => total + source[outcome],
0,
);
}
function assertCount(value: number, label: string, maximum: number): void {
if (!Number.isSafeInteger(value) || value < 0 || value > maximum) {
throw new RangeError(`${label} is outside its startup budget`);
}
}
function assessment(
summary: LegacyShadowStartupSummary,
): LegacyShadowStartupAssessment {
if (summary.ambiguous > 0 || summary.skipped > 0 || summary.failed > 0) {
return 'attention_required';
}
if (summary.remaining) return 'incomplete';
if (summary.pending > 0) return 'waiting_external_callback';
return 'converged';
}
export function createLegacyShadowStartupDifferenceReport(
request: LegacyShadowStartupRequest,
summary: LegacyShadowStartupSummary,
): LegacyShadowStartupDifferenceReport {
const expectedBudget = BUDGETS[request.profile];
if (
request.pageSize !== expectedBudget.pageSize ||
request.maxPages !== expectedBudget.maxPages
) {
throw new RangeError('startup report budget does not match its Profile');
}
const configuredOrigins = [...new Set(request.origins)];
if (
configuredOrigins.length < 1 ||
configuredOrigins.length > MAX_CONFIGURED_LEGACY_SHADOW_ORIGINS
) {
throw new RangeError(
'configured Shadow origin count is outside its budget',
);
}
const maxCandidates = request.pageSize * request.maxPages;
assertCount(summary.pages, 'pages', request.maxPages);
if (summary.pages < 1) {
throw new RangeError('pages must include the executed startup page');
}
assertCount(summary.scanned, 'scanned', request.pageSize * summary.pages);
for (const outcome of LEGACY_SHADOW_STARTUP_OUTCOMES) {
assertCount(summary[outcome], outcome, summary.scanned);
}
if (outcomeTotal(summary) !== summary.scanned) {
throw new RangeError('startup outcomes do not conserve scanned candidates');
}
if (summary.remaining !== (summary.stopReason !== 'complete')) {
throw new RangeError('startup remaining and stop reason disagree');
}
if (
!['complete', 'page_limit', 'cursor_stalled'].includes(summary.stopReason)
) {
throw new RangeError('startup stop reason is invalid');
}
if (summary.stopReason === 'page_limit' && summary.nextCursor === undefined) {
throw new RangeError('page-limited startup summary has no resume cursor');
}
if (summary.stopReason !== 'page_limit' && summary.nextCursor !== undefined) {
throw new RangeError('resume cursor is present without a page limit');
}
if (summary.byOrigin.length !== configuredOrigins.length) {
throw new RangeError('origin outcome coverage is incomplete');
}
const origins = new Map<ExecutionOrigin, LegacyShadowStartupOriginSummary>();
for (const origin of summary.byOrigin) {
if (
origins.has(origin.origin) ||
!configuredOrigins.includes(origin.origin)
) {
throw new RangeError('origin outcome coverage is invalid');
}
assertCount(origin.scanned, `${origin.origin}:scanned`, summary.scanned);
for (const outcome of LEGACY_SHADOW_STARTUP_OUTCOMES) {
assertCount(
origin[outcome],
`${origin.origin}:${outcome}`,
origin.scanned,
);
}
if (outcomeTotal(origin) !== origin.scanned) {
throw new RangeError(
'origin outcomes do not conserve scanned candidates',
);
}
origins.set(origin.origin, {
origin: origin.origin,
scanned: origin.scanned,
...outcomeCounts(origin),
});
}
const orderedOrigins = configuredOrigins.map(
(origin) => origins.get(origin)!,
);
if (
orderedOrigins.reduce((total, origin) => total + origin.scanned, 0) !==
summary.scanned ||
LEGACY_SHADOW_STARTUP_OUTCOMES.some(
(outcome) =>
orderedOrigins.reduce((total, origin) => total + origin[outcome], 0) !==
summary[outcome],
)
) {
throw new RangeError('origin outcomes do not match aggregate outcomes');
}
return {
schemaVersion: 1,
schema: LEGACY_SHADOW_STARTUP_DIFFERENCE_REPORT_SCHEMA,
profile: request.profile,
assessment: assessment(summary),
configuredOriginCount: configuredOrigins.length,
budget: {
pageSize: request.pageSize,
maxPages: request.maxPages,
maxCandidates,
},
coverage: {
pages: summary.pages,
scanned: summary.scanned,
stopReason: summary.stopReason,
remaining: summary.remaining,
resumeAvailable: summary.nextCursor !== undefined,
},
outcomes: outcomeCounts(summary),
byOrigin: orderedOrigins,
};
}
export function createLegacyShadowStartupMetricBatch(
report: LegacyShadowStartupDifferenceReport,
): LegacyShadowStartupMetricBatch {
return {
schemaVersion: 1,
schema: LEGACY_SHADOW_STARTUP_METRIC_BATCH_SCHEMA,
dimensions: {
profile: report.profile,
assessment: report.assessment,
stopReason: report.coverage.stopReason,
},
values: {
configuredOrigins: report.configuredOriginCount,
pageSize: report.budget.pageSize,
maxPages: report.budget.maxPages,
maxCandidates: report.budget.maxCandidates,
pages: report.coverage.pages,
scanned: report.coverage.scanned,
remaining: report.coverage.remaining ? 1 : 0,
resumeAvailable: report.coverage.resumeAvailable ? 1 : 0,
...outcomeCounts(report.outcomes),
},
byOrigin: report.byOrigin.map((origin) => ({
origin: origin.origin,
scanned: origin.scanned,
...outcomeCounts(origin),
})),
};
}
async function executeDefault(
request: LegacyShadowStartupRequest,
): Promise<LegacyShadowStartupSummary> {
@@ -137,22 +377,32 @@ export async function bootstrapLegacyShadowStartupReconciliation(
}
return record;
}
const summary = await (options.execute ?? executeDefault)({
const request: LegacyShadowStartupRequest = {
origins,
profile,
...BUDGETS[profile],
});
};
const summary = await (options.execute ?? executeDefault)(request);
const report = createLegacyShadowStartupDifferenceReport(request, summary);
const metrics = createLegacyShadowStartupMetricBatch(report);
const record: LegacyShadowStartupAudit = {
state: summary.remaining ? 'incomplete' : 'reconciled',
state: report.assessment === 'converged' ? 'reconciled' : 'incomplete',
profile,
origins: origins.length,
summary: auditSummary(summary),
report,
metrics,
};
try {
await audit(record);
} catch {
// Shadow audit output must not affect Legacy startup.
}
try {
await options.collect?.(metrics);
} catch {
// Shadow metric collection must not affect Legacy startup.
}
return record;
} catch (error) {
const record: LegacyShadowStartupAudit = {
@@ -23,41 +23,86 @@ export interface LegacyShadowStartupClock {
now(): number;
}
export interface LegacyShadowStartupReconcileSummary {
scanned: number;
completed: number;
cancelled: number;
abandoned: number;
markedLost: number;
repaired: number;
pending: number;
ambiguous: number;
skipped: number;
failed: number;
truncated: boolean;
nextCursor?: LegacyShadowStartupCursor;
}
export const LEGACY_SHADOW_STARTUP_OUTCOMES = Object.freeze([
'completed',
'cancelled',
'abandoned',
'markedLost',
'repaired',
'pending',
'ambiguous',
'skipped',
'failed',
] as const);
export type LegacyShadowStartupOutcome =
(typeof LEGACY_SHADOW_STARTUP_OUTCOMES)[number];
export type LegacyShadowStartupOutcomeCounts = Record<
LegacyShadowStartupOutcome,
number
>;
export type LegacyShadowStartupOriginSummary =
LegacyShadowStartupOutcomeCounts & {
origin: RunRecord['executionOrigin'];
scanned: number;
};
export type LegacyShadowStartupReconcileSummary =
LegacyShadowStartupOutcomeCounts & {
scanned: number;
byOrigin: readonly LegacyShadowStartupOriginSummary[];
truncated: boolean;
nextCursor?: LegacyShadowStartupCursor;
};
export type LegacyShadowStartupStopReason =
| 'complete'
| 'page_limit'
| 'cursor_stalled';
export interface LegacyShadowStartupSummary
extends Omit<
LegacyShadowStartupReconcileSummary,
'truncated' | 'nextCursor'
> {
export type LegacyShadowStartupSummary = Omit<
LegacyShadowStartupReconcileSummary,
'truncated' | 'nextCursor'
> & {
pages: number;
stopReason: LegacyShadowStartupStopReason;
remaining: boolean;
nextCursor?: LegacyShadowStartupCursor;
}
};
type EvidenceSelection =
| { status: 'matched'; evidence: LegacyRunningInstanceEvidence }
| { status: 'none' | 'ambiguous' };
function emptyOutcomeCounts(): LegacyShadowStartupOutcomeCounts {
return {
completed: 0,
cancelled: 0,
abandoned: 0,
markedLost: 0,
repaired: 0,
pending: 0,
ambiguous: 0,
skipped: 0,
failed: 0,
};
}
function originSummary(
origin: RunRecord['executionOrigin'],
): LegacyShadowStartupOriginSummary {
return { origin, scanned: 0, ...emptyOutcomeCounts() };
}
function incrementOutcome(
target: LegacyShadowStartupOutcomeCounts,
outcome: LegacyShadowStartupOutcome,
): void {
target[outcome] += 1;
}
function sameCursor(
left: LegacyShadowStartupCursor | undefined,
right: LegacyShadowStartupCursor,
@@ -168,27 +213,34 @@ export class LegacyShadowStartupReconciler {
...(options.cursor === undefined ? {} : { cursor: options.cursor }),
...(options.limit === undefined ? {} : { limit: options.limit }),
});
const originSummaries = new Map(
enabledOrigins.map((origin) => [origin, originSummary(origin)]),
);
const summary: LegacyShadowStartupReconcileSummary = {
scanned: page.candidates.length,
completed: 0,
cancelled: 0,
abandoned: 0,
markedLost: 0,
repaired: 0,
pending: 0,
ambiguous: 0,
skipped: 0,
failed: 0,
...emptyOutcomeCounts(),
byOrigin: [...originSummaries.values()],
truncated: page.truncated,
...(page.nextCursor === undefined ? {} : { nextCursor: page.nextCursor }),
};
const origins = new Set(enabledOrigins);
for (const candidate of page.candidates) {
try {
await this.reconcileCandidate(candidate, origins, summary);
} catch {
summary.failed += 1;
const perOrigin =
originSummaries.get(candidate.origin) ??
originSummary(candidate.origin);
if (!originSummaries.has(candidate.origin)) {
originSummaries.set(candidate.origin, perOrigin);
summary.byOrigin = [...originSummaries.values()];
}
perOrigin.scanned += 1;
let outcome: LegacyShadowStartupOutcome;
try {
outcome = await this.reconcileCandidate(candidate, origins);
} catch {
outcome = 'failed';
}
incrementOutcome(summary, outcome);
incrementOutcome(perOrigin, outcome);
}
return summary;
}
@@ -196,26 +248,23 @@ export class LegacyShadowStartupReconciler {
private async reconcileCandidate(
candidate: LegacyShadowStartupCandidate,
origins: ReadonlySet<RunRecord['executionOrigin']>,
summary: LegacyShadowStartupReconcileSummary,
): Promise<void> {
): Promise<LegacyShadowStartupOutcome> {
const run = await this.repository.findRunById(candidate.runId);
if (
!run ||
run.executionOwner !== 'legacy' ||
!origins.has(run.executionOrigin) ||
run.executionOrigin !== candidate.origin ||
isTerminalRunStatus(run.status) ||
run.status === 'lost'
) {
summary.skipped += 1;
return;
return 'skipped';
}
if (candidate.activeAttemptCount === 0) {
await this.repairTerminalAttempt(run, summary);
return;
return this.repairTerminalAttempt(run);
}
if (candidate.activeAttemptCount !== 1 || !candidate.attempt) {
summary.ambiguous += 1;
return;
return 'ambiguous';
}
const attempt = await this.repository.findAttemptById(
candidate.attempt.attemptId,
@@ -225,8 +274,7 @@ export class LegacyShadowStartupReconciler {
attempt.runId !== run.id ||
isTerminalRunAttemptStatus(attempt.status)
) {
summary.skipped += 1;
return;
return 'skipped';
}
const evidence =
@@ -236,13 +284,11 @@ export class LegacyShadowStartupReconciler {
legacyCronId: candidate.legacyCronId,
});
if (evidence.truncated) {
summary.ambiguous += 1;
return;
return 'ambiguous';
}
const selected = selectEvidence(attempt, evidence.evidence);
if (selected.status === 'ambiguous') {
summary.ambiguous += 1;
return;
return 'ambiguous';
}
if (
selected.status === 'matched' &&
@@ -254,8 +300,7 @@ export class LegacyShadowStartupReconciler {
{ runId: run.id, attemptId: attempt.id },
{ atMs, reason: 'reconcile' },
);
summary.cancelled += 1;
return;
return 'cancelled';
}
if (
selected.evidence.outcome === 'succeeded' ||
@@ -274,35 +319,30 @@ export class LegacyShadowStartupReconciler {
: selected.evidence.exitCode,
},
);
summary.completed += 1;
return;
return 'completed';
}
}
if (run.executionOrigin === 'scheduled_system') {
summary.pending += 1;
return;
return 'pending';
}
if (run.status === 'queued' && attempt.status === 'claimed') {
await this.abandon(run, attempt);
summary.abandoned += 1;
return;
return 'abandoned';
}
if (
(run.status === 'dispatching' || run.status === 'running') &&
['claimed', 'starting', 'running'].includes(attempt.status)
) {
await this.markLost(run, attempt);
summary.markedLost += 1;
return;
return 'markedLost';
}
summary.ambiguous += 1;
return 'ambiguous';
}
private async repairTerminalAttempt(
run: RunRecord,
summary: LegacyShadowStartupReconcileSummary,
): Promise<void> {
): Promise<'repaired' | 'ambiguous'> {
const attempt = await this.repository.findLatestAttemptByRunId(run.id);
const target = attempt ? terminalRunTarget(attempt) : undefined;
if (
@@ -310,8 +350,7 @@ export class LegacyShadowStartupReconciler {
target === undefined ||
!RUN_TRANSITIONS[run.status].includes(target)
) {
summary.ambiguous += 1;
return;
return 'ambiguous';
}
await this.commands.transitionRun({
runId: run.id,
@@ -327,7 +366,7 @@ export class LegacyShadowStartupReconciler {
actor: { type: 'reconciler' },
dedupeKey: `legacy-startup-repair:${attempt.id}:${target}`,
});
summary.repaired += 1;
return 'repaired';
}
private async abandon(
@@ -438,18 +477,17 @@ export class LegacyShadowStartupSupervisor {
);
}
const originSummaries = new Map(
[...new Set(options.origins)].map((origin) => [
origin,
originSummary(origin),
]),
);
const total: LegacyShadowStartupSummary = {
pages: 0,
scanned: 0,
completed: 0,
cancelled: 0,
abandoned: 0,
markedLost: 0,
repaired: 0,
pending: 0,
ambiguous: 0,
skipped: 0,
failed: 0,
...emptyOutcomeCounts(),
byOrigin: [...originSummaries.values()],
stopReason: 'complete',
remaining: false,
};
@@ -471,6 +509,19 @@ export class LegacyShadowStartupSupervisor {
total.ambiguous += page.ambiguous;
total.skipped += page.skipped;
total.failed += page.failed;
for (const pageOrigin of page.byOrigin) {
const aggregate =
originSummaries.get(pageOrigin.origin) ??
originSummary(pageOrigin.origin);
if (!originSummaries.has(pageOrigin.origin)) {
originSummaries.set(pageOrigin.origin, aggregate);
total.byOrigin = [...originSummaries.values()];
}
aggregate.scanned += pageOrigin.scanned;
for (const outcome of LEGACY_SHADOW_STARTUP_OUTCOMES) {
aggregate[outcome] += pageOrigin[outcome];
}
}
if (!page.truncated) return total;
if (!page.nextCursor || sameCursor(cursor, page.nextCursor)) {
total.stopReason = 'cursor_stalled';