Docker creating zombie processes

jan · November 9, 2019, 12:08pm

Yes, I also saw on my server that a few zombie processes were there and wanted to investigate. Sadly did not get to it yet. So it would definitely be great if you could check that out! Honestly, do not know much how a process turns into a zombie (and not even what the definition is).

However, the interesting code would be this:

This code starts the workflow processes:

github.com

n8n-io/n8n/blob/master/packages/cli/src/WorkflowRunner.ts

/* eslint-disable @typescript-eslint/no-unsafe-argument */

/* eslint-disable @typescript-eslint/no-unsafe-member-access */

/* eslint-disable @typescript-eslint/no-shadow */
/* eslint-disable @typescript-eslint/no-unsafe-assignment */

import type { IProcessMessage } from 'n8n-core';
import { WorkflowExecute } from 'n8n-core';

import type {
	ExecutionError,
	IDeferredPromise,
	IExecuteResponsePromiseData,
	IPinData,
	IRun,
	WorkflowExecuteMode,
	WorkflowHooks,
} from 'n8n-workflow';
import {

This file has been truncated. show original

This code gets executed in the new process and executes the new workflow:

github.com

n8n-io/n8n/blob/master/packages/cli/src/WorkflowRunnerProcess.ts

/* eslint-disable @typescript-eslint/no-unsafe-argument */
/* eslint-disable @typescript-eslint/no-unsafe-member-access */
/* eslint-disable @typescript-eslint/no-unsafe-assignment */
/* eslint-disable @typescript-eslint/no-shadow */
/* eslint-disable @typescript-eslint/no-use-before-define */
/* eslint-disable @typescript-eslint/unbound-method */
import 'source-map-support/register';
import 'reflect-metadata';
import { setDefaultResultOrder } from 'dns';

import { Container } from 'typedi';
import type { IProcessMessage } from 'n8n-core';
import { BinaryDataService, WorkflowExecute } from 'n8n-core';

import type {
	ExecutionError,
	IDataObject,
	IExecuteResponsePromiseData,
	IExecuteWorkflowInfo,
	INode,

This file has been truncated. show original

Maybe it would be better to not create a new process and instead start it in a worker thread.
There is some discussion about that going on here:

That could maybe at the same time solve this problem, reduce CPU and memory consumption and reduce the time that it takes new workflows to start to execute (which is currently horribly slow with around 1 second).