How to run node.js app forever when console is closed?


Question

I connect to my remote server via ssh. Then I start my node.js app with Forever. Everything works fine until I close my console window. How to run node.js app FOREVER on my remote server even when I close my connection via ssh? I just want to start an app and shut down my copmputer. My app should be working in the background on my remote server.

1
51
4/28/2011 11:56:54 AM

Accepted Answer

You may also want to consider using the upstart utility. It will allow you to start, stop and restart you node application like a service. Upstart can configured to automatically restart your application if it crashes.

Install upstart:

sudo apt-get install upstart

Create a simple script for your application that will look something like:

#!upstart
description "my app"

start on started mountall
stop on shutdown

# Automatically Respawn:
respawn
respawn limit 99 5

env NODE_ENV=production

exec node /somepath/myapp/app.js >> /var/log/myapp.log 2>&1

Then copy the script file (myapp.conf) to /etc/init and make sure its marked as executable. Your application can then be managed using the following commands:

sudo start myapp
sudo stop myapp
sudo restart myapp
57
6/23/2011 1:51:40 PM

Two answers: One for Windows, one for *nix:

On Windows, you can use the start command to start the process disconnected from your instance of cmd.exe:

start node example.js

On *nix, there are two aspects of this: Disconnecting the process from the console, and making sure it doesn't receive the HUP signal ("hang up"), which most processes (including Node) will respond to by terminating. The former is possibly optional, but the latter is necessary.

Starting disconnected from the console is easy: Usually, you just put an ampersand (&) at the end of the command line:

# Keep reading, don't just grab this and use it
node example.js &

But the above doesn't protect the process from HUP signals. The program may or may not receive HUP when you close the shell (console), depending on a shell option called huponexit. If huponexit is true, the process will receive HUP when the shell exits and will presumably terminate.

huponexit defaults to false on the various Linux variants I've used, and in fact I happily used the above for years until coderjoe and others helped me understand (in a very long comment stream under the answer that may have since been deleted) that I was relying on huponexit being false.

To avoid the possibility that huponexit might be true in your environment, explicitly use nohup. nohup runs the process immune from HUP signals. You use it like this:

nohup node example.js > /dev/null &

or

nohup node example.js > your-desired-filename-or-stream-here &

The redirection is important; if you don't do it, you'll end up with a nohup.out file containing the output from stdout and stderr. (By default, nohup redirects stderr to stdout, and if stdout is outputting to a terminal, it redirects that to nohup.out. nohup also redirects stdin if it's receiving from a terminal, so we don't have to do that. See man nohup or info coreutils 'nohup invocation' for details.)

In general for these things, you want to use a process monitor so that if the process crashes for some reason, the monitor restarts it, but the above does work for simple cases.


Licensed under: CC-BY-SA with attribution
Not affiliated with: Stack Overflow
Icon