n8n-io/n8n · error · Error

Snapshot not found: ${filename} (tried ${filePath})

Error message

Snapshot not found: ${filename} (tried ${filePath})

What it means

Plain `Error('Snapshot not found: <filename> (tried <filePath>)')` thrown at e2e.controller.ts:423 when the resolved heap-snapshot file path (from `heapSnapshotPaths` map or `path.resolve(filename)` in cwd) does not exist via `fs.existsSync`. Test-only controller; serialized as a 500 by the generic handler. The message leaks the absolute path tried.

Source

Thrown at packages/cli/src/controllers/e2e.controller.ts:423

	/**
	 * Download a heap snapshot file as a stream.
	 * The response-helper pipes Readable streams directly to the response.
	 */
	@Get('/heap-snapshot/:filename', { skipAuth: true })
	downloadHeapSnapshot(req: Request) {
		const fs = require('node:fs') as typeof nodeFs;
		const path = require('node:path') as typeof nodePath;

		const filename = path.basename(req.params.filename);
		if (!filename.endsWith('.heapsnapshot')) {
			throw new Error('Invalid file type');
		}

		// Look up the full path stored during POST, or try cwd
		const filePath = this.heapSnapshotPaths.get(filename) ?? path.resolve(filename);
		if (!fs.existsSync(filePath)) {
			throw new Error(`Snapshot not found: ${filename} (tried ${filePath})`);
		}

		return fs.createReadStream(filePath);
	}

	// --- Per-spec backend V8 coverage (DEVP-370) -----------------------------
	// Lets the Playwright coverage fixture attribute this main process's V8
	// coverage to the spec that produced it, so backend source files land in the
	// E2E impact map (today the map is frontend-only). Test-only: the whole
	// controller hard-exits outside E2E (see top of file).
	//
	// Uses `Profiler.getBestEffortCoverage` (NON-draining) + a server-side
	// baseline/delta so it never resets the global V8 counters — the shard-level
	// `NODE_V8_COVERAGE` exit dump (the Codecov report path) is left untouched.
	//
	// REQUIRES the coverage run to be single-worker (it is — coverage project
	// runs `--workers=1`). Precise coverage is process-global, so the start→take
	// window is only attributable to one spec when specs don't run concurrently.

View on GitHub (pinned to 5ac6606e81)

Solutions

1. Ensure the POST `/e2e/heap-snapshot` route ran and registered the filename in `heapSnapshotPaths` before GET.
2. Use the exact filename returned by POST — do not synthesize one.
3. Disable any cleanup that removes `.heapsnapshot` files mid-test-run.
4. Run the test from the same working directory as the controller process.

import { existsSync } from 'node:fs';
function snapshotReady(p: string) { return existsSync(p); }
// in the test: assert snapshotReady(filePath) before GET

When it happens

Trigger: In E2E mode: requesting a `.heapsnapshot` filename that was never recorded by the POST route, was already cleaned up, or whose file lives outside the resolved base path. The map lookup falls back to cwd, so a relative-name mismatch also triggers it.

Common situations: Playwright run ordering (download before POST completed); temp-file cleanup between runs; running the GET from a different cwd than the POST; fixture using a fabricated filename.

Related errors

• Invalid file type
• File not found: ${path}
• Role not found
• Workflow "${workflowId}" not found.
• Version "${versionIdToPublish}" not found for workflow "${wo

AI-assisted analysis of n8n-io/n8n@5ac6606e81 (2026-08-12). Data as JSON: /api/errors/f3f2ce2eeb4afb1c. Report an issue: GitHub.