diff --git a/scripts/__tests__/run-vitest-stable-shard.test.mjs b/scripts/__tests__/run-vitest-stable-shard.test.mjs index 889c5fb066..7d0da8ea99 100644 --- a/scripts/__tests__/run-vitest-stable-shard.test.mjs +++ b/scripts/__tests__/run-vitest-stable-shard.test.mjs @@ -4,8 +4,15 @@ import path from "node:path"; import { fileURLToPath } from "node:url"; import test from "node:test"; +import { + defaultSuiteWeight, + loadShardDurations, + partitionGeneralServerSuites, +} from "../general-server-shard.mjs"; + const repoRoot = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "..", ".."); const script = path.join(repoRoot, "scripts", "run-vitest-stable.mjs"); +const durationsManifest = path.join(repoRoot, "scripts", "general-server-shard-durations.json"); function dryRun(args) { const result = spawnSync(process.execPath, [script, ...args, "--dry-run"], { @@ -61,3 +68,80 @@ test("shard flags are rejected for the parallel workspace groups", () => { const result = dryRun(["--mode", "general", "--group", "general-workspaces-a", "--shard-index", "0", "--shard-count", "3"]); assert.notEqual(result.status, 0, "workspace groups must not accept shard flags"); }); + +test("duration-aware partition balances skewed weights better than round-robin", () => { + // Round-robin puts all three heavy suites on shard 0 (indexes 0, 3, 6). + const files = ["a", "b", "c", "d", "e", "f", "g", "h", "i"]; + const durations = { a: 30000, d: 30000, g: 30000, b: 100, c: 100, e: 100, f: 100, h: 100, i: 100 }; + + const shards = partitionGeneralServerSuites(files, 3, durations); + const totals = shards.map((shard) => shard.totalWeight); + const maxTotal = Math.max(...totals); + const minTotal = Math.min(...totals); + assert.ok( + maxTotal - minTotal <= 200, + `expected near-even shard weights, got ${totals.join(", ")}`, + ); + assert.equal( + shards.flatMap((shard) => shard.files).sort().join(","), + files.join(","), + "partition must cover every file exactly once", + ); +}); + +test("the partition is deterministic for identical inputs", () => { + const files = Array.from({ length: 50 }, (_, index) => `suite-${index}.test.ts`); + const durations = Object.fromEntries(files.map((file, index) => [file, (index * 37) % 5000])); + + const first = partitionGeneralServerSuites(files, 3, durations); + const second = partitionGeneralServerSuites(files, 3, durations); + assert.deepEqual(first, second, "same inputs must always produce the same partition"); +}); + +test("suites missing from the manifest get the median weight", () => { + assert.equal(defaultSuiteWeight({ a: 100, b: 300, c: 900 }), 300); + assert.equal(defaultSuiteWeight({ a: 100, b: 300, c: 500, d: 900 }), 400); + assert.equal(defaultSuiteWeight({}), 1000, "empty manifest falls back to a fixed weight"); +}); + +test("a missing or malformed manifest degrades to uniform weights", () => { + assert.deepEqual(loadShardDurations(path.join(repoRoot, "scripts", "no-such-manifest.json")), {}); + + const files = ["a", "b", "c", "d"]; + const shards = partitionGeneralServerSuites(files, 2, {}); + assert.equal(shards[0].files.length + shards[1].files.length, files.length); + assert.equal(Math.abs(shards[0].files.length - shards[1].files.length), 0); +}); + +test("the checked-in manifest loads and covers most of the current suite set", () => { + const durations = loadShardDurations(durationsManifest); + assert.ok(Object.keys(durations).length > 0, "manifest must parse to a non-empty duration map"); + + const shard = dryRunJson(["--mode", "general", "--group", "general-server", "--shard-index", "0", "--shard-count", "1"]); + const currentFiles = shard.selectedGeneralServerSuites; + const known = currentFiles.filter((file) => durations[file] !== undefined).length; + assert.ok( + known / currentFiles.length >= 0.5, + `manifest is stale: only ${known} of ${currentFiles.length} suites have recorded durations — regenerate it from a recent PR run (see the manifest's $comment)`, + ); +}); + +test("the real shard partition is duration-balanced", () => { + const durations = loadShardDurations(durationsManifest); + const fallback = defaultSuiteWeight(durations); + const shards = Array.from({ length: SHARD_COUNT }, (_, index) => + dryRunJson(["--mode", "general", "--group", "general-server", "--shard-index", String(index), "--shard-count", String(SHARD_COUNT)]), + ); + + const totals = shards.map((shard) => + shard.selectedGeneralServerSuites.reduce((sum, file) => sum + (durations[file] ?? fallback), 0), + ); + const maxTotal = Math.max(...totals); + const minTotal = Math.min(...totals); + // LPT keeps the spread within the heaviest single suite; use that as the bound. + const heaviest = Math.max(...Object.values(durations)); + assert.ok( + maxTotal - minTotal <= heaviest, + `shard weight spread ${maxTotal - minTotal}ms exceeds heaviest suite ${heaviest}ms: ${totals.join(", ")}`, + ); +}); diff --git a/scripts/general-server-shard-durations.json b/scripts/general-server-shard-durations.json new file mode 100644 index 0000000000..8b73e9318c --- /dev/null +++ b/scripts/general-server-shard-durations.json @@ -0,0 +1,246 @@ +{ + "$comment": "Per-suite wall-clock durations (ms) for the general-server vitest lane, used by scripts/general-server-shard.mjs to balance suites across the PR shard matrix. Sampled from a real PR run of .github/workflows/pr.yml (actions run 29262068391, 2026-07-13). Suites missing here get the median weight, so the manifest only needs occasional refreshes: pull the 'Run grouped general test suites' logs from a recent PR run and record each suite's vitest-reported duration.", + "unit": "ms", + "durations": { + "server/src/__tests__/access-service.test.ts": 2303, + "server/src/__tests__/access-validators.test.ts": 7, + "server/src/__tests__/activity-service.test.ts": 2390, + "server/src/__tests__/adapter-models.test.ts": 61, + "server/src/__tests__/adapter-registry.test.ts": 17, + "server/src/__tests__/adapter-session-codecs.test.ts": 10, + "server/src/__tests__/agent-auth-jwt.test.ts": 20, + "server/src/__tests__/agent-auth-middleware.test.ts": 66, + "server/src/__tests__/agent-instructions-service.test.ts": 58, + "server/src/__tests__/agent-invokability.test.ts": 5, + "server/src/__tests__/agent-permissions-service.test.ts": 6, + "server/src/__tests__/agent-shortname-collision.test.ts": 5, + "server/src/__tests__/agent-skill-contract.test.ts": 6, + "server/src/__tests__/agents-pending-approval-config.test.ts": 2182, + "server/src/__tests__/agents-service-clear-error.test.ts": 2258, + "server/src/__tests__/agents-service-secret-bindings.test.ts": 2367, + "server/src/__tests__/api-compression.test.ts": 77, + "server/src/__tests__/app-hmr-port.test.ts": 4, + "server/src/__tests__/app-private-hostname-gate.test.ts": 4, + "server/src/__tests__/app-vite-dev-routing.test.ts": 4, + "server/src/__tests__/approvals-service.test.ts": 12, + "server/src/__tests__/attachment-types.test.ts": 9, + "server/src/__tests__/attention-service.test.ts": 2626, + "server/src/__tests__/authorization-service.test.ts": 2824, + "server/src/__tests__/aws-secrets-manager-provider.test.ts": 39, + "server/src/__tests__/better-auth.test.ts": 9, + "server/src/__tests__/board-claim.test.ts": 2121, + "server/src/__tests__/board-mutation-guard.test.ts": 50, + "server/src/__tests__/body-limits.test.ts": 3, + "server/src/__tests__/budgets-service.test.ts": 2211, + "server/src/__tests__/built-in-agents.test.ts": 6157, + "server/src/__tests__/change-consent-gate.test.ts": 2256, + "server/src/__tests__/claude-local-adapter-environment.test.ts": 372, + "server/src/__tests__/claude-local-adapter.test.ts": 8, + "server/src/__tests__/claude-local-execute.test.ts": 2771, + "server/src/__tests__/claude-local-skill-sync.test.ts": 21, + "server/src/__tests__/cleanup-removal-service.test.ts": 2248, + "server/src/__tests__/cloud-upstreams.test.ts": 2576, + "server/src/__tests__/codex-auth-reconciliation.test.ts": 46, + "server/src/__tests__/codex-local-adapter.test.ts": 9, + "server/src/__tests__/codex-local-execute.test.ts": 974, + "server/src/__tests__/codex-local-skill-injection.test.ts": 29, + "server/src/__tests__/codex-local-skill-sync.test.ts": 18, + "server/src/__tests__/companies-service.test.ts": 3104, + "server/src/__tests__/company-artifacts-service.test.ts": 2531, + "server/src/__tests__/company-search-service.test.ts": 3127, + "server/src/__tests__/company-skill-test-runs-service.test.ts": 3331, + "server/src/__tests__/company-skills-catalog-service.test.ts": 3584, + "server/src/__tests__/company-skills-detail.test.ts": 2258, + "server/src/__tests__/company-skills-service.test.ts": 5470, + "server/src/__tests__/company-skills.test.ts": 33, + "server/src/__tests__/cursor-local-adapter-environment.test.ts": 232, + "server/src/__tests__/cursor-local-adapter.test.ts": 11, + "server/src/__tests__/cursor-local-execute.test.ts": 1028, + "server/src/__tests__/cursor-local-skill-injection.test.ts": 15, + "server/src/__tests__/cursor-local-skill-sync.test.ts": 22, + "server/src/__tests__/dashboard-service.test.ts": 2222, + "server/src/__tests__/dev-runner-output.test.ts": 25, + "server/src/__tests__/dev-runner-paths.test.ts": 4, + "server/src/__tests__/dev-runner-snapshot.test.ts": 8, + "server/src/__tests__/dev-runner-worktree.test.ts": 7, + "server/src/__tests__/dev-server-status.test.ts": 7, + "server/src/__tests__/dev-watch-ignore.test.ts": 7, + "server/src/__tests__/docker-entrypoint.test.ts": 39, + "server/src/__tests__/document-annotations-service.test.ts": 2528, + "server/src/__tests__/documents-service.test.ts": 2253, + "server/src/__tests__/documents.test.ts": 4, + "server/src/__tests__/effective-run-config-fingerprints.test.ts": 10, + "server/src/__tests__/environment-config.test.ts": 11, + "server/src/__tests__/environment-custom-image-terminal-ws.test.ts": 2566, + "server/src/__tests__/environment-custom-images-service.test.ts": 3889, + "server/src/__tests__/environment-execution-target.test.ts": 7, + "server/src/__tests__/environment-live-ssh.test.ts": 717, + "server/src/__tests__/environment-probe.test.ts": 12, + "server/src/__tests__/environment-run-orchestrator.test.ts": 13, + "server/src/__tests__/environment-runtime-driver-contract.test.ts": 2723, + "server/src/__tests__/environment-runtime.test.ts": 3287, + "server/src/__tests__/environment-service.test.ts": 2266, + "server/src/__tests__/environment-test-harness.test.ts": 14, + "server/src/__tests__/error-handler.test.ts": 12, + "server/src/__tests__/execution-lock-orphan-cleanup.test.ts": 2333, + "server/src/__tests__/execution-workspace-policy.test.ts": 9, + "server/src/__tests__/execution-workspaces-service.test.ts": 6035, + "server/src/__tests__/external-objects-service.test.ts": 2282, + "server/src/__tests__/feedback-flush-controller.test.ts": 4, + "server/src/__tests__/feedback-service.test.ts": 2520, + "server/src/__tests__/feedback-share-client.test.ts": 9, + "server/src/__tests__/file-resources.test.ts": 4757, + "server/src/__tests__/first-admin-claim.test.ts": 2117, + "server/src/__tests__/forbidden-tokens.test.ts": 6, + "server/src/__tests__/gemini-local-adapter-environment.test.ts": 80, + "server/src/__tests__/gemini-local-adapter.test.ts": 9, + "server/src/__tests__/gemini-local-execute.test.ts": 220, + "server/src/__tests__/gemini-local-skill-sync.test.ts": 18, + "server/src/__tests__/grok-local-skill-sync.test.ts": 15, + "server/src/__tests__/heartbeat-accepted-plan-workspace-refresh.test.ts": 5034, + "server/src/__tests__/heartbeat-active-run-output-watchdog.test.ts": 49231, + "server/src/__tests__/heartbeat-archived-company-guard.test.ts": 2303, + "server/src/__tests__/heartbeat-auto-checkout.test.ts": 5, + "server/src/__tests__/heartbeat-comment-wake-batching.test.ts": 5030, + "server/src/__tests__/heartbeat-context-summary.test.ts": 8, + "server/src/__tests__/heartbeat-issue-rewake-throttle.test.ts": 3198, + "server/src/__tests__/heartbeat-list.test.ts": 2240, + "server/src/__tests__/heartbeat-local-environment.test.ts": 2504, + "server/src/__tests__/heartbeat-lock-release-on-reassignment.test.ts": 2129, + "server/src/__tests__/heartbeat-model-profile.test.ts": 6, + "server/src/__tests__/heartbeat-plugin-environment.test.ts": 2310, + "server/src/__tests__/heartbeat-project-env.test.ts": 24, + "server/src/__tests__/heartbeat-responsible-user-invariant.test.ts": 6285, + "server/src/__tests__/heartbeat-retry-scheduling.test.ts": 3373, + "server/src/__tests__/heartbeat-run-log.test.ts": 5, + "server/src/__tests__/heartbeat-run-summary.test.ts": 6, + "server/src/__tests__/heartbeat-runtime-skills.test.ts": 2681, + "server/src/__tests__/heartbeat-runtime-state.test.ts": 2322, + "server/src/__tests__/heartbeat-scheduling-suppression.test.ts": 5, + "server/src/__tests__/heartbeat-stale-queue-invalidation.test.ts": 11573, + "server/src/__tests__/heartbeat-start-lock.test.ts": 8, + "server/src/__tests__/heartbeat-timer-wake-session-reset-pf4.test.ts": 6, + "server/src/__tests__/heartbeat-workspace-branch-containment.test.ts": 6144, + "server/src/__tests__/heartbeat-workspace-finalize-branch.test.ts": 3599, + "server/src/__tests__/heartbeat-workspace-session.test.ts": 185, + "server/src/__tests__/heartbeat-worktree-suppression.test.ts": 2665, + "server/src/__tests__/heartbeat-zombie-guard.test.ts": 5, + "server/src/__tests__/hire-hook.test.ts": 14, + "server/src/__tests__/http-log-policy.test.ts": 5, + "server/src/__tests__/inbox-dismissals.test.ts": 2110, + "server/src/__tests__/instance-settings-service.test.ts": 15, + "server/src/__tests__/instrumentation.test.ts": 27, + "server/src/__tests__/invite-join-grants.test.ts": 6, + "server/src/__tests__/issue-blocker-attention.test.ts": 3329, + "server/src/__tests__/issue-comment-redaction.test.ts": 2383, + "server/src/__tests__/issue-continuation-summary.test.ts": 6, + "server/src/__tests__/issue-execution-policy.test.ts": 35, + "server/src/__tests__/issue-goal-fallback.test.ts": 5, + "server/src/__tests__/issue-liveness.test.ts": 14, + "server/src/__tests__/issue-monitor-scheduler.test.ts": 6810, + "server/src/__tests__/issue-recovery-actions.test.ts": 3556, + "server/src/__tests__/issue-references-service.test.ts": 2184, + "server/src/__tests__/issue-rewake-throttle.test.ts": 7, + "server/src/__tests__/issue-thread-interactions-service.test.ts": 3053, + "server/src/__tests__/issue-thread-interactions-telemetry.test.ts": 2209, + "server/src/__tests__/issue-tree-control-service-unit.test.ts": 6, + "server/src/__tests__/issue-tree-control-service.test.ts": 2498, + "server/src/__tests__/issues-list-query-parsing.test.ts": 42, + "server/src/__tests__/issues-user-context.test.ts": 5, + "server/src/__tests__/join-request-dedupe.test.ts": 4, + "server/src/__tests__/json-schema-secret-refs.test.ts": 4, + "server/src/__tests__/live-events-ws.test.ts": 21, + "server/src/__tests__/log-redaction.test.ts": 6, + "server/src/__tests__/logger-tz.test.ts": 24, + "server/src/__tests__/monthly-spend-service.test.ts": 29, + "server/src/__tests__/openclaw-gateway-adapter.test.ts": 52, + "server/src/__tests__/opencode-local-adapter.test.ts": 8, + "server/src/__tests__/opencode-local-skill-sync.test.ts": 17, + "server/src/__tests__/paperclip-env.test.ts": 5, + "server/src/__tests__/paperclip-skill-utils.test.ts": 18, + "server/src/__tests__/parse-status-filter.test.ts": 6, + "server/src/__tests__/pi-local-adapter-environment.test.ts": 128, + "server/src/__tests__/pi-local-execute.test.ts": 228, + "server/src/__tests__/pi-local-skill-sync.test.ts": 17, + "server/src/__tests__/pipelines-service.test.ts": 4478, + "server/src/__tests__/plugin-access-authorization-host-services.test.ts": 2259, + "server/src/__tests__/plugin-database.test.ts": 2422, + "server/src/__tests__/plugin-dev-watcher.test.ts": 67, + "server/src/__tests__/plugin-environment-driver-seam.test.ts": 55, + "server/src/__tests__/plugin-execution-workspace-bridge.test.ts": 7, + "server/src/__tests__/plugin-install-autobuild.test.ts": 4336, + "server/src/__tests__/plugin-lifecycle-restart.test.ts": 7, + "server/src/__tests__/plugin-local-folders.test.ts": 63, + "server/src/__tests__/plugin-managed-agents.test.ts": 2407, + "server/src/__tests__/plugin-managed-routines.test.ts": 2371, + "server/src/__tests__/plugin-managed-skills.test.ts": 2600, + "server/src/__tests__/plugin-orchestration-apis.test.ts": 2531, + "server/src/__tests__/plugin-sdk-orchestration-contract.test.ts": 13, + "server/src/__tests__/plugin-sdk-testing.test.ts": 10, + "server/src/__tests__/plugin-secrets-handler.test.ts": 5, + "server/src/__tests__/plugin-telemetry-bridge.test.ts": 36, + "server/src/__tests__/plugin-tenant-isolation.test.ts": 2264, + "server/src/__tests__/plugin-tool-dispatcher-pluginDbId.test.ts": 12, + "server/src/__tests__/plugin-worker-manager.test.ts": 355, + "server/src/__tests__/private-hostname-guard.test.ts": 39, + "server/src/__tests__/productivity-review-service.test.ts": 5180, + "server/src/__tests__/project-icon-persistence.test.ts": 2091, + "server/src/__tests__/project-list-metrics.test.ts": 5, + "server/src/__tests__/project-shortname-resolution.test.ts": 5, + "server/src/__tests__/qa-routine-secrets-e2e.test.ts": 2364, + "server/src/__tests__/quota-windows-service.test.ts": 7, + "server/src/__tests__/quota-windows.test.ts": 59, + "server/src/__tests__/recovery-classifiers.test.ts": 9, + "server/src/__tests__/recovery-stale-issue-lock-sweep.test.ts": 2250, + "server/src/__tests__/redact-sensitive.test.ts": 7, + "server/src/__tests__/routine-run-telemetry.test.ts": 2157, + "server/src/__tests__/routines-service.test.ts": 5411, + "server/src/__tests__/run-continuations.test.ts": 7, + "server/src/__tests__/run-liveness.test.ts": 11, + "server/src/__tests__/runtime-api.test.ts": 6, + "server/src/__tests__/sandbox-provider-runtime.test.ts": 9, + "server/src/__tests__/secret-provider-registry.test.ts": 6, + "server/src/__tests__/secrets-service.test.ts": 3281, + "server/src/__tests__/server-info.test.ts": 300, + "server/src/__tests__/server-package-build-script.test.ts": 3, + "server/src/__tests__/server-startup-feedback-export.test.ts": 27, + "server/src/__tests__/shared-telemetry-events.test.ts": 8, + "server/src/__tests__/skills-catalog-service.test.ts": 39, + "server/src/__tests__/source-trust.test.ts": 2204, + "server/src/__tests__/static-index-html.test.ts": 28, + "server/src/__tests__/storage-local-provider.test.ts": 23, + "server/src/__tests__/task-watchdogs-classifier.test.ts": 9, + "server/src/__tests__/task-watchdogs-scheduler.test.ts": 3245, + "server/src/__tests__/teams-catalog-install-no-overrides.test.ts": 2995, + "server/src/__tests__/teams-catalog-service.test.ts": 75, + "server/src/__tests__/telemetry-client-flush.test.ts": 12, + "server/src/__tests__/trust-preset-resolver.test.ts": 10, + "server/src/__tests__/trust-proxy.test.ts": 11, + "server/src/__tests__/ui-branding.test.ts": 6, + "server/src/__tests__/version.test.ts": 11, + "server/src/__tests__/vite-html-renderer.test.ts": 6, + "server/src/__tests__/work-products.test.ts": 6, + "server/src/__tests__/work-timeline-service.test.ts": 2541, + "server/src/__tests__/workspace-runtime.test.ts": 28685, + "server/src/__tests__/worktree-config.test.ts": 29, + "server/src/adapters/claude-agent-id-header.test.ts": 6, + "server/src/adapters/http/execute.test.ts": 7, + "server/src/middleware/cloud-tenant-actor.test.ts": 10, + "server/src/services/adapter-models-env.test.ts": 5, + "server/src/services/adapter-registry-bootstrap.reconcile.test.ts": 7, + "server/src/services/adapter-registry-bootstrap.test.ts": 7, + "server/src/services/environment-custom-image-terminal-sessions.test.ts": 11, + "server/src/services/execution-allowlist.test.ts": 6, + "server/src/services/execution-policy-bootstrap.test.ts": 15, + "server/src/services/heartbeat-run-runtime-status.test.ts": 9, + "server/src/services/heartbeat-stop-metadata.test.ts": 6, + "server/src/services/issue-thread-interactions.test.ts": 757, + "server/src/services/recovery/model-profile-hint.test.ts": 4, + "server/src/services/recovery/service.pause-durability.test.ts": 4, + "server/src/services/recovery/successful-run-handoff.test.ts": 10, + "server/src/services/responsible-user-denial-run-outcomes.test.ts": 8, + "server/src/services/routines-formatter-cache.test.ts": 1029, + "server/src/services/run-scratch.test.ts": 16, + "server/src/services/session-workspace-cwd.test.ts": 4 + } +} diff --git a/scripts/general-server-shard.mjs b/scripts/general-server-shard.mjs new file mode 100644 index 0000000000..16ff2165f9 --- /dev/null +++ b/scripts/general-server-shard.mjs @@ -0,0 +1,75 @@ +import { readFileSync } from "node:fs"; + +// Fallback weight (ms) when the duration manifest is missing or empty. +const FALLBACK_SUITE_WEIGHT_MS = 1000; + +// Loads the per-suite duration manifest produced from a real PR run (see the +// $comment field in scripts/general-server-shard-durations.json). Returns an +// empty map on any read/parse problem so sharding degrades to uniform weights +// instead of failing the test lane. +export function loadShardDurations(manifestPath) { + let parsed; + try { + parsed = JSON.parse(readFileSync(manifestPath, "utf8")); + } catch { + return {}; + } + + const durations = parsed?.durations; + if (!durations || typeof durations !== "object" || Array.isArray(durations)) { + return {}; + } + + const result = {}; + for (const [file, ms] of Object.entries(durations)) { + if (typeof ms === "number" && Number.isFinite(ms) && ms >= 0) { + result[file] = ms; + } + } + return result; +} + +// Weight assigned to suites absent from the manifest (new or renamed files). +// The median keeps one unknown suite from skewing a shard the way a mean +// dragged up by a few 30s+ suites would. +export function defaultSuiteWeight(durations) { + const values = Object.values(durations).sort((a, b) => a - b); + if (values.length === 0) { + return FALLBACK_SUITE_WEIGHT_MS; + } + const mid = Math.floor(values.length / 2); + return values.length % 2 === 1 ? values[mid] : (values[mid - 1] + values[mid]) / 2; +} + +// Deterministic longest-processing-time partition: heaviest suite first, each +// assigned to the currently lightest shard. Ties break by file path and then +// by shard index, so every runner in the matrix computes the identical +// partition from the same checkout — that invariant is what makes the shards +// a complete, non-overlapping cover of the suite set. +export function partitionGeneralServerSuites(files, shardCount, durations = {}) { + const fallbackWeight = defaultSuiteWeight(durations); + const weighted = files + .map((file) => ({ file, weight: durations[file] ?? fallbackWeight })) + .sort((a, b) => b.weight - a.weight || a.file.localeCompare(b.file)); + + const shards = Array.from({ length: shardCount }, () => ({ files: [], totalWeight: 0 })); + for (const { file, weight } of weighted) { + let target = 0; + for (let index = 1; index < shards.length; index += 1) { + if (shards[index].totalWeight < shards[target].totalWeight) { + target = index; + } + } + shards[target].files.push(file); + shards[target].totalWeight += weight; + } + + for (const shard of shards) { + shard.files.sort((a, b) => a.localeCompare(b)); + } + return shards; +} + +export function selectGeneralServerShard(files, shardIndex, shardCount, durations = {}) { + return partitionGeneralServerSuites(files, shardCount, durations)[shardIndex].files; +} diff --git a/scripts/run-vitest-stable.mjs b/scripts/run-vitest-stable.mjs index 69cca3b188..d37ee4d405 100644 --- a/scripts/run-vitest-stable.mjs +++ b/scripts/run-vitest-stable.mjs @@ -3,8 +3,14 @@ import { spawnSync } from "node:child_process"; import { mkdirSync, mkdtempSync, readdirSync, statSync } from "node:fs"; import os from "node:os"; import path from "node:path"; +import { fileURLToPath } from "node:url"; +import { loadShardDurations, selectGeneralServerShard } from "./general-server-shard.mjs"; const repoRoot = process.cwd(); +const scriptsDir = path.dirname(fileURLToPath(import.meta.url)); +const generalServerShardDurations = loadShardDurations( + path.join(scriptsDir, "general-server-shard-durations.json"), +); const serverRoot = path.join(repoRoot, "server"); const serverSrcDir = path.join(repoRoot, "server", "src"); const serverTestsDir = path.join(repoRoot, "server", "src", "__tests__"); @@ -288,8 +294,11 @@ function runProjectGroup(projects, groupName) { function runGeneralGroup(routeTests, groupName, shardIndex = null, shardCount = null) { if (groupName === generalServerGroupName) { if (shardCount !== null && shardCount > 1) { - const shardFiles = generalServerTestFiles.filter( - (_, index) => index % shardCount === shardIndex, + const shardFiles = selectGeneralServerShard( + generalServerTestFiles, + shardIndex, + shardCount, + generalServerShardDurations, ); console.log( `\n[test:run] general-server shard ${shardIndex + 1}/${shardCount} running ${shardFiles.length} of ${generalServerTestFiles.length} suites`, @@ -369,6 +378,8 @@ const routeTests = walk(serverTestsDir) // dedicated serialized shards. Sharding this list across runners is what keeps // the general-server lane from becoming the PR critical path: the server vitest // config pins maxWorkers to 1, so the only way to parallelize is across jobs. +// Suites are partitioned by recorded duration (scripts/general-server-shard.mjs) +// rather than round-robin, so one slow suite cluster can't stretch a single shard. const generalServerTestFiles = walk(serverSrcDir) .map((file) => toRepoPath(file)) .filter((repoPath) => repoPath.endsWith(".test.ts")) @@ -396,8 +407,11 @@ if (options.dryRun) { options.mode === generalModeName && options.group === generalServerGroupName && options.shardCount !== null - ? generalServerTestFiles.filter( - (_, index) => index % options.shardCount === options.shardIndex, + ? selectGeneralServerShard( + generalServerTestFiles, + options.shardIndex, + options.shardCount, + generalServerShardDurations, ) : null, }, diff --git a/server/src/__tests__/heartbeat-issue-rewake-throttle.test.ts b/server/src/__tests__/heartbeat-issue-rewake-throttle.test.ts index 637512f399..2d044de81c 100644 --- a/server/src/__tests__/heartbeat-issue-rewake-throttle.test.ts +++ b/server/src/__tests__/heartbeat-issue-rewake-throttle.test.ts @@ -74,19 +74,31 @@ describeEmbeddedPostgres("heartbeat issue rewake throttle", () => { if (!runs.some((run) => run.status === "queued" || run.status === "running")) break; await new Promise((resolve) => setTimeout(resolve, 50)); } - await db.delete(environmentLeases); - await db.delete(issueComments); - await db.delete(issues); - await db.delete(heartbeatRunEvents); - await db.delete(activityLog); - await db.delete(heartbeatRuns); - await db.delete(agentWakeupRequests); - await db.delete(agentRuntimeState); - await db.delete(agents); - await db.delete(environments); - await db.delete(executionWorkspaces); - await db.delete(companySkills); - await db.delete(companies); + // Post-run bookkeeping (run-event records, follow-up wake scheduling) can + // still write for a moment after a run reaches a terminal status, so a + // single delete sweep can hit a foreign-key violation when a late insert + // lands between two deletes. Retry the sweep until it goes through clean. + for (let attempt = 0; ; attempt += 1) { + try { + await db.delete(environmentLeases); + await db.delete(issueComments); + await db.delete(issues); + await db.delete(heartbeatRunEvents); + await db.delete(activityLog); + await db.delete(heartbeatRuns); + await db.delete(agentWakeupRequests); + await db.delete(agentRuntimeState); + await db.delete(agents); + await db.delete(environments); + await db.delete(executionWorkspaces); + await db.delete(companySkills); + await db.delete(companies); + break; + } catch (error) { + if (attempt >= 4) throw error; + await new Promise((resolve) => setTimeout(resolve, 100)); + } + } }); afterAll(async () => {