n8n/packages/testing/playwright/playwright-projects.ts
Declan Carroll 3a33a448b0
Some checks failed
CI: Master (Build, Test, Lint) / Build for Github Cache (push) Has been cancelled
CI: Master (Build, Test, Lint) / Unit tests (22.x) (push) Has been cancelled
CI: Master (Build, Test, Lint) / Unit tests (24.14.1) (push) Has been cancelled
CI: Master (Build, Test, Lint) / Unit tests (25.x) (push) Has been cancelled
CI: Master (Build, Test, Lint) / Lint (push) Has been cancelled
CI: Master (Build, Test, Lint) / Performance (push) Has been cancelled
CI: Master (Build, Test, Lint) / Notify Slack on failure (push) Has been cancelled
Util: Update Node Popularity / update-popularity (push) Has been cancelled
Test: E2E Coverage Weekly / Prepare Docker (coverage) (push) Has been cancelled
Util: Update Node Popularity / approve-and-automerge (push) Has been cancelled
Test: E2E Coverage Weekly / E2E (coverage) (push) Has been cancelled
Test: E2E Coverage Weekly / Aggregate Coverage (push) Has been cancelled
Release: Schedule Patch Release PRs / Create patch release PR (${{ matrix.track }}) (beta) (push) Has been cancelled
Release: Schedule Patch Release PRs / Create patch release PR (${{ matrix.track }}) (stable) (push) Has been cancelled
Release: Schedule Patch Release PRs / Create patch release PR (${{ matrix.track }}) (v1) (push) Has been cancelled
test(benchmark): Question-driven Playwright benchmark suite with tiered topology and rich diagnostics (#29024)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-09 21:14:08 +00:00

299 lines
10 KiB
TypeScript

import type { Project } from '@playwright/test';
import type { N8NConfig } from 'n8n-containers/stack';
import {
CONTAINER_ONLY_CAPABILITIES,
CONTAINER_ONLY_MODES,
LICENSED_TAG,
} from './fixtures/capabilities';
import { getBackendUrl, getFrontendUrl } from './utils/url-helper';
// Tests that require container environment (won't run against local n8n).
// Matches:
// - @capability:X - add-on features (email, proxy, source-control, etc.)
// - @mode:X - infrastructure modes (postgres, queue, multi-main)
// - @licensed - enterprise license features (log streaming, SSO, etc.)
// - @db:reset - tests needing per-test database reset (requires isolated containers)
const CONTAINER_ONLY = new RegExp(
[
`@capability:(${CONTAINER_ONLY_CAPABILITIES.join('|')})`,
`@mode:(${CONTAINER_ONLY_MODES.join('|')})`,
`@${LICENSED_TAG}`,
'@db:reset',
].join('|'),
);
// Escape hatch: allow `@capability:*` tests to run against a pre-started local
// n8n. Fixtures that depend on container-provided services (proxy, mailpit,
// etc.) must detect the no-container case and skip or fall back to direct
// network calls. Used by `pnpm test:local:isolated` and similar workflows.
const ALLOW_CONTAINER_ONLY = process.env.PLAYWRIGHT_ALLOW_CONTAINER_ONLY === 'true';
const CONTAINER_CONFIGS: Array<{ name: string; config: N8NConfig }> = [
{ name: 'sqlite', config: {} },
{ name: 'postgres', config: { postgres: true } },
{ name: 'queue', config: { workers: 1 } },
{
name: 'multi-main',
config: { mains: 2, workers: 1, services: ['victoriaLogs', 'victoriaMetrics', 'vector'] },
},
];
// --- Benchmark profiles ---
// Each profile represents a real-world n8n deployment configuration.
// ONE test file runs in ALL profiles — adding a profile auto-expands coverage.
// Standard benchmark resource profile.
// Main: 2 vCPU, 4 GB RAM
// Worker: 1 vCPU, 2 GB RAM
// Total host budget at 1m+3w: 5 vCPU + 10 GB for n8n, plus ~3 GB for
// postgres/kafka/redis/observability.
export const BENCHMARK_MAIN_RESOURCES = { memory: 4, cpu: 2 };
export const BENCHMARK_WORKER_RESOURCES = { memory: 2, cpu: 1 };
export const OBSERVABILITY_SERVICES = ['victoriaLogs', 'victoriaMetrics', 'vector'] as const;
/**
* Single benchmark stack config. Specs call `benchConfig()` to get a copy with
* spec-specific overrides (mains, workers, kafka). All n8n env tuning lives
* here once — the queue-mode-only vars (`QUEUE_*`, `DB_PING_INTERVAL_SECONDS`)
* are inert in direct mode, so a single env profile works for both.
*/
const BENCHMARK_CONFIG: N8NConfig = {
// Postgres exporter scrapes DB internals into VictoriaMetrics; cAdvisor exposes per-container
// CPU/memory/IO so benchmarks can detect when PG/n8n hit OS-level limits the per-query
// reporter alone would miss. Only meaningful for benchmarks.
services: [...OBSERVABILITY_SERVICES, 'postgresExporter', 'cadvisor'],
postgres: true,
resourceQuota: BENCHMARK_MAIN_RESOURCES,
workerResourceQuota: BENCHMARK_WORKER_RESOURCES,
// Distribute load across all mains. UI tests stick to the default `first`
// policy so debugging hits a single predictable backend.
lbPolicy: 'round_robin',
env: {
N8N_LOG_LEVEL: 'error',
N8N_DIAGNOSTICS_ENABLED: 'false',
N8N_METRICS_INCLUDE_MESSAGE_EVENT_BUS_METRICS: 'true',
N8N_METRICS_INCLUDE_QUEUE_METRICS: 'true',
DB_POSTGRESDB_POOL_SIZE: '10',
DB_POSTGRESDB_CONNECTION_TIMEOUT: '300000',
DB_PING_INTERVAL_SECONDS: '5',
N8N_CONCURRENCY_PRODUCTION_LIMIT: '20',
QUEUE_BULL_REDIS_KEEP_ALIVE: 'true',
QUEUE_BULL_REDIS_TIMEOUT_THRESHOLD: '60000',
QUEUE_WORKER_LOCK_DURATION: '300000',
QUEUE_WORKER_LOCK_RENEW_TIME: '20000',
QUEUE_WORKER_STALLED_INTERVAL: '60000',
QUEUE_RECOVERY_INTERVAL: '0',
EXECUTIONS_DATA_SAVE_ON_SUCCESS: 'none',
},
};
export interface BenchOptions {
/** Adds the `kafka` service. Default: false. */
kafka?: boolean;
/** Number of main pods. Default: 1. Multi-main HA env enabled when > 1. */
mains?: number;
/** Number of worker pods. Default: 0 (direct mode). */
workers?: number;
/**
* Adds the `tracing` service (Jaeger + n8n-tracer) and turns on OTEL emission.
* Adds ~5-10% per-request overhead — opt in only when measuring OTEL cost or
* collecting flamegraph data, not for clean ceiling numbers.
*/
tracing?: boolean;
/** Additional env vars to merge over the base. */
env?: Record<string, string>;
}
/**
* Build a benchmark stack config. Each spec calls this with a unique
* `isolation` slug (which becomes `TEST_ISOLATION` so each spec gets its own
* container). Pass topology overrides via `opts`.
*
* @example
* // Direct-mode kafka (no workers)
* test.use({ capability: benchConfig('single-instance-ceiling', { kafka: true }) });
*
* // Queue-mode kafka (1 main + 3 workers)
* test.use({ capability: benchConfig('node-count-scaling', { kafka: true, workers: 3 }) });
*
* // Multi-main webhook
* test.use({ capability: benchConfig('webhook-main-scaling', { mains: 2, workers: 2 }) });
*/
export function benchConfig(isolation: string, opts: BenchOptions = {}): N8NConfig {
const services = [...(BENCHMARK_CONFIG.services ?? [])];
if (opts.kafka) services.push('kafka');
if (opts.tracing) services.push('tracing');
const env: Record<string, string> = {
...BENCHMARK_CONFIG.env,
...(opts.tracing && {
N8N_OTEL_ENABLED: 'true',
N8N_OTEL_EXPORTER_OTLP_ENDPOINT: 'http://jaeger:4318',
N8N_OTEL_EXPORTER_SERVICE_NAME: `n8n-bench-${isolation}`,
N8N_OTEL_TRACES_INCLUDE_NODE_SPANS: 'true',
}),
...opts.env,
TEST_ISOLATION: `bench-${isolation}`,
};
if ((opts.mains ?? 1) > 1) env.N8N_MULTI_MAIN_SETUP_ENABLED = 'true';
return {
...BENCHMARK_CONFIG,
services,
...(opts.mains !== undefined && { mains: opts.mains }),
...(opts.workers !== undefined && { workers: opts.workers }),
env,
};
}
type BenchmarkProfile = { name: string; config: N8NConfig };
// Project-level fallback config for benchmark specs that don't call
// `benchConfig()` themselves. In practice all current specs do, so this just
// has to be a valid stack — content doesn't matter.
const BENCHMARKING_DEFAULT_CONFIG: N8NConfig = benchConfig('default', { kafka: true });
// Benchmark profiles that host local-only tests (model API keys, long runtimes,
// reserved metric names). They scan `tests/infrastructure/benchmarks-local/` and
// are intentionally left out of the CI matrix.
const LOCAL_ONLY_BENCHMARK_PROFILES: BenchmarkProfile[] = [
{
name: 'memory-instanceai',
config: {
...BENCHMARK_CONFIG,
services: [...OBSERVABILITY_SERVICES],
env: {
...BENCHMARK_CONFIG.env,
// Instance-AI module & model config
N8N_ENABLED_MODULES: 'instance-ai',
N8N_INSTANCE_AI_MODEL: process.env.N8N_INSTANCE_AI_MODEL ?? 'openai/gpt-4o-mini',
// Forward API keys to the container
...(process.env.N8N_AI_OPENAI_API_KEY && {
N8N_AI_OPENAI_API_KEY: process.env.N8N_AI_OPENAI_API_KEY,
OPENAI_API_KEY: process.env.N8N_AI_OPENAI_API_KEY,
}),
...(process.env.LANGSMITH_API_KEY && {
LANGSMITH_API_KEY: process.env.LANGSMITH_API_KEY,
}),
...(process.env.ANTHROPIC_API_KEY && {
ANTHROPIC_API_KEY: process.env.ANTHROPIC_API_KEY,
}),
...(process.env.OPENROUTER_API_KEY && {
OPENROUTER_API_KEY: process.env.OPENROUTER_API_KEY,
}),
...(process.env.CEREBRAS_API_KEY && {
CEREBRAS_API_KEY: process.env.CEREBRAS_API_KEY,
}),
// Sandbox config — forwarded from host env if present
...(process.env.N8N_INSTANCE_AI_SANDBOX_ENABLED && {
N8N_INSTANCE_AI_SANDBOX_ENABLED: process.env.N8N_INSTANCE_AI_SANDBOX_ENABLED,
N8N_INSTANCE_AI_SANDBOX_PROVIDER:
process.env.N8N_INSTANCE_AI_SANDBOX_PROVIDER ?? 'daytona',
N8N_INSTANCE_AI_SANDBOX_IMAGE: process.env.N8N_INSTANCE_AI_SANDBOX_IMAGE ?? '',
}),
...(process.env.DAYTONA_API_URL && {
DAYTONA_API_URL: process.env.DAYTONA_API_URL,
}),
...(process.env.DAYTONA_API_KEY && {
DAYTONA_API_KEY: process.env.DAYTONA_API_KEY,
}),
},
},
},
];
export function getProjects(): Project[] {
const isLocal = !!getBackendUrl();
const projects: Project[] = [];
if (isLocal) {
projects.push({
name: 'e2e',
testDir: './tests/e2e',
grepInvert: ALLOW_CONTAINER_ONLY ? undefined : CONTAINER_ONLY,
fullyParallel: true,
use: { baseURL: getFrontendUrl() },
});
projects.push({
name: 'dev-server-smoke',
testDir: './tests/dev-server-smoke',
fullyParallel: false,
// Vite dev cold-start can take 15-25s on the first navigation while it
// pre-bundles deps. Subsequent navigations are sub-second.
timeout: 90_000,
use: { baseURL: getFrontendUrl(), navigationTimeout: 30_000 },
});
} else {
for (const { name, config } of CONTAINER_CONFIGS) {
projects.push(
{
name: `${name}:e2e`,
testDir: './tests/e2e',
timeout: name === 'sqlite' ? 60000 : 180000, // 60 seconds for sqlite container test, 180 for other modes to allow startup
fullyParallel: true,
use: { containerConfig: config },
},
{
name: `${name}:infrastructure`,
testDir: './tests/infrastructure',
grep: new RegExp(`@mode:${name}|@capability:${name}`),
workers: 1,
timeout: 180000,
use: { containerConfig: config },
},
);
}
projects.push({
name: 'coverage',
testDir: './tests/e2e',
timeout: 60000,
fullyParallel: true,
use: { containerConfig: {} },
});
projects.push({
name: 'benchmarking:infrastructure',
testDir: './tests/infrastructure/benchmarks',
workers: 1,
timeout: 600_000,
retries: 0,
use: { containerConfig: BENCHMARKING_DEFAULT_CONFIG },
});
for (const { name, config } of LOCAL_ONLY_BENCHMARK_PROFILES) {
projects.push({
name: `benchmark-${name}:infrastructure`,
testDir: './tests/infrastructure/benchmarks-local',
workers: 1,
timeout: 600_000,
retries: 0,
use: { containerConfig: config },
});
}
}
projects.push({
name: 'cli-workflows',
testDir: './tests/cli-workflows',
fullyParallel: true,
timeout: 60000,
});
projects.push({
name: 'performance',
testDir: './tests/performance',
workers: 1,
timeout: 300000,
retries: 0,
use: {
// Default container config for performance tests, equivalent to @cloud:starter
containerConfig: { resourceQuota: { memory: 0.75, cpu: 0.5 }, env: { E2E_TESTS: 'true' } },
},
});
return projects;
}