wehub-resource-sync
508 lines
21 KiB
TypeScript
508 lines
21 KiB
TypeScript
import { describe, expect, onTestFinished, vi } from "vitest";
|
|
|
|
// db.server + splitMode are mocked so the idempotency dedup client resolves to
|
|
// the container prisma passed into the concern (split stays off).
|
|
vi.mock("~/db.server", () => ({
|
|
prisma: {},
|
|
$replica: {},
|
|
runOpsNewPrisma: {},
|
|
runOpsLegacyPrisma: {},
|
|
}));
|
|
|
|
vi.mock("~/v3/runOpsMigration/splitMode.server", () => ({ isSplitEnabled: async () => false }));
|
|
|
|
vi.mock("~/services/platform.v3.server", async (importOriginal) => {
|
|
const actual = (await importOriginal()) as Record<string, unknown>;
|
|
return {
|
|
...actual,
|
|
getEntitlement: vi.fn(),
|
|
};
|
|
});
|
|
|
|
import { RunEngine } from "@internal/run-engine";
|
|
import { setupAuthenticatedEnvironment, setupBackgroundWorker } from "@internal/run-engine/tests";
|
|
import { containerTest } from "@internal/testcontainers";
|
|
import { trace } from "@opentelemetry/api";
|
|
import { IdempotencyKeyConcern } from "~/runEngine/concerns/idempotencyKeys.server";
|
|
import { DefaultQueueManager } from "~/runEngine/concerns/queues.server";
|
|
import type { ValidationResult } from "~/runEngine/types";
|
|
import { RunEngineTriggerTaskService } from "../../app/runEngine/services/triggerTask.server";
|
|
import {
|
|
MOCK_SPAN_ID,
|
|
MOCK_TRACE_ID,
|
|
MockPayloadProcessor,
|
|
MockTraceEventConcern,
|
|
MockTriggerTaskValidator,
|
|
} from "./triggerTaskTestHelpers";
|
|
|
|
vi.setConfig({ testTimeout: 60_000, hookTimeout: 60_000 });
|
|
|
|
describe("RunEngineTriggerTaskService", () => {
|
|
// ─── Mollifier integration ──────────────────────────────────────────────────
|
|
//
|
|
// These tests pin the call-site behaviour of the mollifier hooks inside
|
|
// RunEngineTriggerTaskService.call. They use the optional DI ports
|
|
// (`evaluateGate`, `getMollifierBuffer`) added on the service constructor —
|
|
// production wiring is unchanged (defaults to the live module-level imports).
|
|
// Each test's regression intent lives in its own setup comment.
|
|
|
|
class CapturingMollifierBuffer {
|
|
public accepted: Array<{ runId: string; envId: string; orgId: string; payload: string }> = [];
|
|
async accept(input: { runId: string; envId: string; orgId: string; payload: string }) {
|
|
this.accepted.push(input);
|
|
return true;
|
|
}
|
|
async pop() {
|
|
return null;
|
|
}
|
|
async ack() {}
|
|
async requeue() {}
|
|
async fail() {
|
|
return false;
|
|
}
|
|
async getEntry() {
|
|
return null;
|
|
}
|
|
async listEnvs(): Promise<string[]> {
|
|
return [];
|
|
}
|
|
async getEntryTtlSeconds(): Promise<number> {
|
|
return -1;
|
|
}
|
|
async evaluateTrip() {
|
|
return { tripped: false, count: 0 };
|
|
}
|
|
async close() {}
|
|
}
|
|
|
|
containerTest(
|
|
"mollifier · validation throws before the gate is consulted; no buffer write",
|
|
async ({ prisma, redisOptions }) => {
|
|
const engine = new RunEngine({
|
|
prisma,
|
|
worker: { redis: redisOptions, workers: 1, tasksPerWorker: 10, pollIntervalMs: 100 },
|
|
queue: { redis: redisOptions },
|
|
runLock: { redis: redisOptions },
|
|
machines: {
|
|
defaultMachine: "small-1x",
|
|
machines: {
|
|
"small-1x": { name: "small-1x" as const, cpu: 0.5, memory: 0.5, centsPerMs: 0.0001 },
|
|
},
|
|
baseCostInCents: 0.0005,
|
|
},
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
});
|
|
onTestFinished(() => engine.quit());
|
|
|
|
const authenticatedEnvironment = await setupAuthenticatedEnvironment(prisma, "PRODUCTION");
|
|
const taskIdentifier = "test-task";
|
|
await setupBackgroundWorker(engine, authenticatedEnvironment, taskIdentifier);
|
|
|
|
// Validator that fails on maxAttempts. Any validation throw must abort
|
|
// the call BEFORE the gate runs — otherwise the gate could leak a
|
|
// buffer write for an invalid request.
|
|
class FailingMaxAttemptsValidator extends MockTriggerTaskValidator {
|
|
validateMaxAttempts(): ValidationResult {
|
|
return { ok: false, error: new Error("synthetic max-attempts failure") };
|
|
}
|
|
}
|
|
|
|
const buffer = new CapturingMollifierBuffer();
|
|
const evaluateGateSpy = vi.fn(async () => ({
|
|
action: "mollify" as const,
|
|
decision: {
|
|
divert: true as const,
|
|
reason: "per_env_rate" as const,
|
|
count: 99,
|
|
threshold: 1,
|
|
windowMs: 200,
|
|
holdMs: 500,
|
|
},
|
|
}));
|
|
|
|
const triggerTaskService = new RunEngineTriggerTaskService({
|
|
engine,
|
|
prisma,
|
|
payloadProcessor: new MockPayloadProcessor(),
|
|
queueConcern: new DefaultQueueManager(prisma, engine),
|
|
idempotencyKeyConcern: new IdempotencyKeyConcern(
|
|
prisma,
|
|
engine,
|
|
new MockTraceEventConcern()
|
|
),
|
|
validator: new FailingMaxAttemptsValidator(),
|
|
traceEventConcern: new MockTraceEventConcern(),
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
metadataMaximumSize: 1024 * 1024,
|
|
evaluateGate: evaluateGateSpy,
|
|
getMollifierBuffer: () => buffer as never,
|
|
isMollifierGloballyEnabled: () => true,
|
|
});
|
|
|
|
await expect(
|
|
triggerTaskService.call({
|
|
taskId: taskIdentifier,
|
|
environment: authenticatedEnvironment,
|
|
body: { payload: { test: "x" } },
|
|
})
|
|
).rejects.toThrow(/synthetic max-attempts failure/);
|
|
|
|
// Critical: the gate must NEVER be consulted when validation fails.
|
|
// If this assertion fires, validation has been re-ordered after the
|
|
// mollifier gate — a regression that would let invalid triggers land
|
|
// in the buffer.
|
|
expect(evaluateGateSpy).not.toHaveBeenCalled();
|
|
expect(buffer.accepted).toHaveLength(0);
|
|
}
|
|
);
|
|
|
|
containerTest(
|
|
"mollifier · mollify action writes to buffer and returns synthetic result (no Postgres row)",
|
|
async ({ prisma, redisOptions }) => {
|
|
// When the gate decides mollify, the call site
|
|
// invokes `mollifyTrigger` which writes the engine.trigger snapshot
|
|
// to the buffer and returns a synthesised `MollifySyntheticResult`
|
|
// (run.friendlyId + notice + isCached:false). `engine.trigger` is
|
|
// NEVER invoked on this path — the run materialises in Postgres
|
|
// later, when the drainer replays the snapshot. The replay is
|
|
// covered by `mollifierDrainerHandler.test.ts`; this test pins the
|
|
// call-site integration: synthetic result + buffer write + no
|
|
// Postgres side effect.
|
|
const engine = new RunEngine({
|
|
prisma,
|
|
worker: { redis: redisOptions, workers: 1, tasksPerWorker: 10, pollIntervalMs: 100 },
|
|
queue: { redis: redisOptions },
|
|
runLock: { redis: redisOptions },
|
|
machines: {
|
|
defaultMachine: "small-1x",
|
|
machines: {
|
|
"small-1x": { name: "small-1x" as const, cpu: 0.5, memory: 0.5, centsPerMs: 0.0001 },
|
|
},
|
|
baseCostInCents: 0.0005,
|
|
},
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
});
|
|
onTestFinished(() => engine.quit());
|
|
|
|
const authenticatedEnvironment = await setupAuthenticatedEnvironment(prisma, "PRODUCTION");
|
|
const taskIdentifier = "test-task";
|
|
await setupBackgroundWorker(engine, authenticatedEnvironment, taskIdentifier);
|
|
|
|
// Buffer override records the time of the accept call so we can
|
|
// assert that traceRun fired strictly before the buffer was
|
|
// touched. If a future change re-introduces the "skip traceRun on
|
|
// mollify" shortcut, traceConcern.traceRunEnteredAt stays
|
|
// undefined and the ordering assertion fails.
|
|
class TimestampedBuffer extends CapturingMollifierBuffer {
|
|
public acceptedAt: number | undefined;
|
|
override async accept(input: {
|
|
runId: string;
|
|
envId: string;
|
|
orgId: string;
|
|
payload: string;
|
|
}) {
|
|
this.acceptedAt = Date.now();
|
|
return await super.accept(input);
|
|
}
|
|
}
|
|
const buffer = new TimestampedBuffer();
|
|
const trippedDecision = {
|
|
divert: true as const,
|
|
reason: "per_env_rate" as const,
|
|
count: 150,
|
|
threshold: 100,
|
|
windowMs: 200,
|
|
holdMs: 500,
|
|
};
|
|
const traceConcern = new MockTraceEventConcern();
|
|
|
|
const triggerTaskService = new RunEngineTriggerTaskService({
|
|
engine,
|
|
prisma,
|
|
payloadProcessor: new MockPayloadProcessor(),
|
|
queueConcern: new DefaultQueueManager(prisma, engine),
|
|
idempotencyKeyConcern: new IdempotencyKeyConcern(
|
|
prisma,
|
|
engine,
|
|
new MockTraceEventConcern()
|
|
),
|
|
validator: new MockTriggerTaskValidator(),
|
|
traceEventConcern: traceConcern,
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
metadataMaximumSize: 1024 * 1024,
|
|
evaluateGate: async () => ({ action: "mollify", decision: trippedDecision }),
|
|
getMollifierBuffer: () => buffer as never,
|
|
isMollifierGloballyEnabled: () => true,
|
|
});
|
|
|
|
const result = await triggerTaskService.call({
|
|
taskId: taskIdentifier,
|
|
environment: authenticatedEnvironment,
|
|
body: { payload: { hello: "world" } },
|
|
});
|
|
|
|
// Pre-modifier span creation: traceRun must run BEFORE the buffer
|
|
// is touched. Customer-visible effect — the run span lands in
|
|
// ClickHouse from the moment the trigger returns, even when the
|
|
// drainer is offline, so buffered runs are visible in the trace
|
|
// view immediately rather than only after drain.
|
|
expect(traceConcern.traceRunEnteredAt).toBeDefined();
|
|
expect(buffer.acceptedAt).toBeDefined();
|
|
expect(traceConcern.traceRunEnteredAt!).toBeLessThanOrEqual(buffer.acceptedAt!);
|
|
|
|
// Synthetic result is returned with the `mollifier.queued` notice
|
|
// (the call-site casts the synthetic shape to `TriggerTaskServiceResult`;
|
|
// at runtime the `notice` and `isCached: false` fields are present
|
|
// and read by the api.v1.tasks.$taskId.trigger.ts route handler).
|
|
expect(result).toBeDefined();
|
|
expect(result?.run.friendlyId).toBeDefined();
|
|
const synthetic = result as unknown as {
|
|
run: { friendlyId: string };
|
|
isCached: false;
|
|
notice: { code: string; message: string; docs: string };
|
|
};
|
|
expect(synthetic.isCached).toBe(false);
|
|
expect(synthetic.notice.code).toBe("mollifier.queued");
|
|
expect(synthetic.notice.message).toBeTypeOf("string");
|
|
expect(synthetic.notice.docs).toBeTypeOf("string");
|
|
|
|
// The mollify branch must flag `isMollified: true` on the result so
|
|
// the trigger route can skip `saveRequestIdempotency`. Caching the
|
|
// synthetic runId in the request-idempotency table would mean a
|
|
// lost-response SDK retry (same `x-trigger-request-idempotency-key`
|
|
// header) hits a PG miss in `handleRequestIdempotency` and falls
|
|
// through to a fresh trigger — producing a duplicate buffer entry
|
|
// for trigger calls without a task-level idempotency key. The
|
|
// bounded behaviour (accept retry-as-fresh-trigger during the
|
|
// buffer window) is the deliberate choice; a stale-cache lookup
|
|
// returning null is not.
|
|
expect(result?.isMollified).toBe(true);
|
|
|
|
// buffer.accept ran — Redis has the canonical engine.trigger snapshot
|
|
// under the synthesised friendlyId. The drainer will read this and
|
|
// replay it through engine.trigger to materialise the run.
|
|
expect(buffer.accepted).toHaveLength(1);
|
|
expect(buffer.accepted[0]!.runId).toBe(result!.run.friendlyId);
|
|
expect(buffer.accepted[0]!.envId).toBe(authenticatedEnvironment.id);
|
|
expect(buffer.accepted[0]!.orgId).toBe(authenticatedEnvironment.organizationId);
|
|
// Payload is a JSON-serialised MollifierSnapshot (the engine.trigger
|
|
// input). Schema is internal to the engine, so we only assert that
|
|
// it parses and references the friendlyId — anything more specific
|
|
// would couple the mollifier-layer test to engine-layer fields.
|
|
const snapshot = JSON.parse(buffer.accepted[0]!.payload) as {
|
|
traceId?: string;
|
|
spanId?: string;
|
|
traceContext?: { traceparent?: string };
|
|
};
|
|
|
|
// Regression guard for the dashboard trace-tree bug: the mollifier
|
|
// snapshot MUST carry a W3C `traceparent` in `traceContext`,
|
|
// seeded from the same span traceRun opened. Without it, the
|
|
// drainer replays through engine.trigger with empty traceContext
|
|
// and every downstream `recordRunDebugLog`
|
|
// (QUEUED/EXECUTING/FINISHED/run:notify…) gets a fresh traceId +
|
|
// null parentId — the run-detail page can only show the root
|
|
// span. Both the mollify and pass-through paths now flow through
|
|
// `traceEventConcern.traceRun`; this assertion pins the
|
|
// seeding-from-the-run-span contract.
|
|
expect(snapshot.traceContext?.traceparent).toMatch(
|
|
/^00-[0-9a-f]{32}-[0-9a-f]{16}-[0-9a-f]{2}$/
|
|
);
|
|
expect(snapshot.traceContext!.traceparent).toContain(snapshot.traceId);
|
|
expect(snapshot.traceContext!.traceparent).toContain(snapshot.spanId);
|
|
// The snapshot inherits the *run span's* traceId/spanId (from the
|
|
// event handed in by traceRun), not a separately-generated OTel
|
|
// span. This is what lets the drainer's `mollifier.drained` span
|
|
// and downstream engine.trigger materialisation parent on the
|
|
// same ClickHouse trace the customer sees from the moment trigger
|
|
// returns.
|
|
expect(snapshot.traceId).toBe(MOCK_TRACE_ID);
|
|
expect(snapshot.spanId).toBe(MOCK_SPAN_ID);
|
|
|
|
// Postgres has NOT been written: engine.trigger was never called on
|
|
// the mollify path. The run materialises only when the drainer
|
|
// replays the snapshot. Regression intent: if a future change makes
|
|
// the mollify branch fall through to engine.trigger (re-introducing
|
|
// phase-1 dual-write), this assertion fails loudly.
|
|
const pgRun = await prisma.taskRun.findFirst({
|
|
where: { friendlyId: result!.run.friendlyId },
|
|
});
|
|
expect(pgRun).toBeNull();
|
|
}
|
|
);
|
|
|
|
containerTest(
|
|
"mollifier · pass_through action does NOT call buffer.accept",
|
|
async ({ prisma, redisOptions }) => {
|
|
const engine = new RunEngine({
|
|
prisma,
|
|
worker: { redis: redisOptions, workers: 1, tasksPerWorker: 10, pollIntervalMs: 100 },
|
|
queue: { redis: redisOptions },
|
|
runLock: { redis: redisOptions },
|
|
machines: {
|
|
defaultMachine: "small-1x",
|
|
machines: {
|
|
"small-1x": { name: "small-1x" as const, cpu: 0.5, memory: 0.5, centsPerMs: 0.0001 },
|
|
},
|
|
baseCostInCents: 0.0005,
|
|
},
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
});
|
|
onTestFinished(() => engine.quit());
|
|
|
|
const authenticatedEnvironment = await setupAuthenticatedEnvironment(prisma, "PRODUCTION");
|
|
const taskIdentifier = "test-task";
|
|
await setupBackgroundWorker(engine, authenticatedEnvironment, taskIdentifier);
|
|
|
|
const buffer = new CapturingMollifierBuffer();
|
|
const getBufferSpy = vi.fn(() => buffer as never);
|
|
|
|
const triggerTaskService = new RunEngineTriggerTaskService({
|
|
engine,
|
|
prisma,
|
|
payloadProcessor: new MockPayloadProcessor(),
|
|
queueConcern: new DefaultQueueManager(prisma, engine),
|
|
idempotencyKeyConcern: new IdempotencyKeyConcern(
|
|
prisma,
|
|
engine,
|
|
new MockTraceEventConcern()
|
|
),
|
|
validator: new MockTriggerTaskValidator(),
|
|
traceEventConcern: new MockTraceEventConcern(),
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
metadataMaximumSize: 1024 * 1024,
|
|
evaluateGate: async () => ({ action: "pass_through" }),
|
|
getMollifierBuffer: getBufferSpy,
|
|
isMollifierGloballyEnabled: () => true,
|
|
});
|
|
|
|
const result = await triggerTaskService.call({
|
|
taskId: taskIdentifier,
|
|
environment: authenticatedEnvironment,
|
|
body: { payload: { test: "x" } },
|
|
});
|
|
|
|
expect(result).toBeDefined();
|
|
// Postgres has the run, no buffer side-effects
|
|
expect(buffer.accepted).toHaveLength(0);
|
|
// getMollifierBuffer must not be called either — the call site short-circuits
|
|
// before touching the singleton when the gate says pass_through.
|
|
expect(getBufferSpy).not.toHaveBeenCalled();
|
|
// Pass-through must NOT set `isMollified` — `result.run` is a real
|
|
// PG row, and the trigger route's `saveRequestIdempotency` is
|
|
// safe to call. Setting the flag here would silently skip the
|
|
// request-idempotency cache for every non-mollified trigger on a
|
|
// mollifier-enabled org, breaking lost-response retry dedup.
|
|
expect(result?.isMollified).toBeFalsy();
|
|
}
|
|
);
|
|
|
|
containerTest(
|
|
"mollifier · idempotency-key match short-circuits BEFORE the gate is consulted",
|
|
async ({ prisma, redisOptions }) => {
|
|
// SCENARIO: a trigger arrives with an idempotency key matching an
|
|
// already-created run. `IdempotencyKeyConcern.handleTriggerRequest`
|
|
// (line 236 of triggerTask.server.ts) detects the match BEFORE the
|
|
// mollifier gate runs and returns `{ isCached: true, run }`. The
|
|
// service early-returns. The gate is never consulted, buffer.accept
|
|
// never fires, no orphan entry is created.
|
|
//
|
|
// Regression intent: if IdempotencyKeyConcern were re-ordered to run
|
|
// AFTER evaluateGate, every idempotent retry on a flagged org would
|
|
// produce an orphan buffer entry — the audit-trail invariant ("every
|
|
// buffered runId has a matching TaskRun") would silently start failing
|
|
// for retries. This test pins the current order.
|
|
|
|
const engine = new RunEngine({
|
|
prisma,
|
|
worker: { redis: redisOptions, workers: 1, tasksPerWorker: 10, pollIntervalMs: 100 },
|
|
queue: { redis: redisOptions },
|
|
runLock: { redis: redisOptions },
|
|
machines: {
|
|
defaultMachine: "small-1x",
|
|
machines: {
|
|
"small-1x": { name: "small-1x" as const, cpu: 0.5, memory: 0.5, centsPerMs: 0.0001 },
|
|
},
|
|
baseCostInCents: 0.0005,
|
|
},
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
});
|
|
onTestFinished(() => engine.quit());
|
|
|
|
const authenticatedEnvironment = await setupAuthenticatedEnvironment(prisma, "PRODUCTION");
|
|
const taskIdentifier = "test-task";
|
|
await setupBackgroundWorker(engine, authenticatedEnvironment, taskIdentifier);
|
|
|
|
const idempotencyKeyConcern = new IdempotencyKeyConcern(
|
|
prisma,
|
|
engine,
|
|
new MockTraceEventConcern()
|
|
);
|
|
|
|
// Setup: normal trigger to create the cached run (no mollifier).
|
|
const baseline = new RunEngineTriggerTaskService({
|
|
engine,
|
|
prisma,
|
|
payloadProcessor: new MockPayloadProcessor(),
|
|
queueConcern: new DefaultQueueManager(prisma, engine),
|
|
idempotencyKeyConcern,
|
|
validator: new MockTriggerTaskValidator(),
|
|
traceEventConcern: new MockTraceEventConcern(),
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
metadataMaximumSize: 1024 * 1024,
|
|
});
|
|
const first = await baseline.call({
|
|
taskId: taskIdentifier,
|
|
environment: authenticatedEnvironment,
|
|
body: { payload: { test: "x" }, options: { idempotencyKey: "regression-key-5" } },
|
|
});
|
|
expect(first?.isCached).toBe(false);
|
|
|
|
// Action: same idempotency key, with a mollify-stub gate that WOULD
|
|
// create an orphan if reached. The concern must short-circuit first.
|
|
const buffer = new CapturingMollifierBuffer();
|
|
const evaluateGateSpy = vi.fn(async () => ({
|
|
action: "mollify" as const,
|
|
decision: {
|
|
divert: true as const,
|
|
reason: "per_env_rate" as const,
|
|
count: 150,
|
|
threshold: 100,
|
|
windowMs: 200,
|
|
holdMs: 500,
|
|
},
|
|
}));
|
|
|
|
const mollifierService = new RunEngineTriggerTaskService({
|
|
engine,
|
|
prisma,
|
|
payloadProcessor: new MockPayloadProcessor(),
|
|
queueConcern: new DefaultQueueManager(prisma, engine),
|
|
idempotencyKeyConcern,
|
|
validator: new MockTriggerTaskValidator(),
|
|
traceEventConcern: new MockTraceEventConcern(),
|
|
tracer: trace.getTracer("test", "0.0.0"),
|
|
metadataMaximumSize: 1024 * 1024,
|
|
evaluateGate: evaluateGateSpy,
|
|
getMollifierBuffer: () => buffer as never,
|
|
isMollifierGloballyEnabled: () => true,
|
|
});
|
|
|
|
const cached = await mollifierService.call({
|
|
taskId: taskIdentifier,
|
|
environment: authenticatedEnvironment,
|
|
body: { payload: { test: "x" }, options: { idempotencyKey: "regression-key-5" } },
|
|
});
|
|
|
|
// Customer sees the cached run, isCached=true
|
|
expect(cached).toBeDefined();
|
|
expect(cached?.isCached).toBe(true);
|
|
expect(cached?.run.friendlyId).toBe(first?.run.friendlyId);
|
|
|
|
// Critical: the gate must NEVER be consulted on a cached-idempotency replay.
|
|
expect(evaluateGateSpy).not.toHaveBeenCalled();
|
|
expect(buffer.accepted).toHaveLength(0);
|
|
}
|
|
);
|
|
});
|