A lightweight, zero-dependency package that wraps your script to restart it once it crashes and to provide other useful stuff
> npm i --save Node-Wrap
Menu: Minimal Example - Full Example - Boot Loop Detection - Forceful KeepAlive - Manual Shutdown / Start - Status Code Table - Disclaimer / Issues / Licensing - TLDR
...whether it is an API of some sort or something like a Discord bot, you will run into these problems:
- Node always exits the script if an error can't be caught
- Even if you don't have that problem, you might still want the ability to restart the entire script
- Maybe you want to add something that requires dynamic shutdowns of your script, like an automatic updater
require("Node-Wrap")("./path/to/the/script/that/should/be/wrapped.js");The above is the smallest possible way to call the wrapper.
First of all, it sets these default settings if none are provided:
restartOnCrash: true, // whether the child process should be restarted after it crashed
crashTimeout: 2000, // the timeout after a crash after which the child process should be restarted
restartTimeout: 0, // the timeout after a restart command after which the child process should be restarted
console: true, // whether Node-Wrap should log some important info to the main console (stuff like "Starting process" and "Restarting process")
logFile: null, // logs all status codes to that file, leave null or undefined for no file logging
logConsoleOutput: null, // logs all console outputs of the child process to that file, leave null or undefined for no file logging
logTimestamp: true, // whether a timestamp should be added to the above logs
restartCodes: [], // what additional exit codes should invoke a restart
bootLoopDetection: 0, // boot loop prevention mechanism: enter the estimated time in milliseconds it usually takes to INITIALIZE (until an infinite loop of some sort gets started) the child process (0 or leave empty to disable) (higher number = higher stability but also longer delay until the boot loop detection kicks in - if you're unsure or it's unstable, take the biggest number of your measurements)
alwaysKeepAlive: false // set to true to force Node-Wrap to insistently keep alive / restart the child process as fast and reliably as possible (unaffected by boot loop detection though)Afterwards the script begins by starting the provided script as a child process.
Because the "logConsoleOutput" property defaulted to null, all stdout (console) output of the child script will be sent to the console of the wrapper.
Now the child process already runs!
To activate the wrapper's magic powers, you need to exit the child process with a certain code from the following list.
| Code | Action | Description |
|---|---|---|
0 |
Stop | This completely stops the child and wrapping process |
1 |
Crash | This triggers a restart if the "restartOnCrash" property of the settings was not set to false (it defaults to true) |
2 |
Restart | This triggers an intentional restart |
3 |
Soft Stop | This stops only the child process but lets Node-Wrap keep running |
So let's say your Discord bot has a developer-only "/restart" command.
To trigger an intentional restart you need to exit the script with the status code 2 (see the table above), just like this:
process.exit(2);...uhh yes, that's it already! You don't need to add anything more to your script(s)!
The boot loop detection mechanism was added in the 0.2.0 update. It basically detects a boot loop and shuts down the script to preserve processing power and bandwidth and free up not needed RAM.
Maybe you just introduced an experimental feature to your heavy, resource intensive API. This feature could break your startup sequence completely, resulting in a boot loop.
With that error (and you for example on vacation for a week or just at work) you could basically DOS yourself and compromise your entire network and other services / APIs.
It's really simple! You just need to measure the time it takes for your script to initialize (aka connect to other services, read / process files / etc. until it reaches an "infinite loop") and add it to the options object like this:
require("Node-Wrap")("./index.js", {
...
bootLoopDetection: 3000 // (in ms) highest number of your measurements (the higher the more stable but also a longer delay until the detection kicks in)
}The above setting will calculate a threshold time for you. If while this threshold is not reached the child process crashes five times in a row, the child process will not be restarted on crash anymore.
You might have to play around with this time a bit but usually the time until your script initialized is the median value.
Node-Wrap has got you covered again! Just add the alwaysKeepAlive property to your options object like this:
require("Node-Wrap")("./index.js", {
...
alwaysKeepAlive: true // this makes Node-Wrap try its hardest to keep the child process alive at all cost
}The above example will force Node-Wrap to restart the child process even if an internal error occurred.
Note: this disables all boot loop detections and automatically reduces all timeouts to 50ms. This means a boot loop could be hogging resources like crazy if it occurred.
So let's say you want your script to automatically update itself.
Achieving this is also made pretty easy with Node-Wrap.
First, check for an update. If one is available, you can shut down the child process, download and install the update and then start the child process again, just like this:
const nodeWrap = require("Node-Wrap"); // include the package after installing it
nodeWrap("./file.js", options); // initialize the file (take a look at the full example to see all available options)
if(updateIsAvailable) updateScript(); // call the function updateScript() if an update is available
function updateScript() {
// Stop the child process but NOT this parent process, so the files can be replaced by the updated ones. Stopping it like this, the child process can be started at a later time again:
nodeWrap.stop(); // this requires the child process to be initialized and started first
downloadAndInstallNewVersion().then(()=>{
// After installing the new version by replacing the old files, start the child process again:
nodeWrap.start(); // this requires the child process to be initialized and manually stopped first
});
}This shutdown behavior can also be triggered from inside the child process, by exiting with the code 3, like this:
process.exit(3);- Create another script file that only contains the minimal example above or the full example at the bottom of this readme if you want to configure it yourself
- Change the "main" property in your "package.json" to start the new file you just made instead (also change other things if you have any to make sure the new file is being run)
- Add the needed
process.exit(code)functions to your script files to trigger the wanted actions (see list of exit codes)
require("Node-Wrap")("./index.js", {
restartOnCrash: true, // whether the child process should be restarted after it crashed
crashTimeout: 20000, // the timeout after a crash after which the child process should be restarted
restartTimeout: 1000, // the timeout after a restart command after which the child process should be restarted
console: true, // whether Node-Wrap should log some important info to the main console (stuff like "Starting process" and "Restarting process")
logFile: "./logs/wrapper.log", // logs all status codes to that file, leave null or undefined for no file logging
logConsoleOutput: "./logs/console.log", // logs all console outputs of the child process to that file, leave null or undefined for no file logging
logTimestamp: false, // whether a timestamp should be added to the above logs
restartCodes: [4, 5, 6], // what additional exit codes should invoke a restart
bootLoopDetection: 3000, // boot loop prevention mechanism: enter the estimated time in milliseconds it usually takes to INITIALIZE (until an infinite loop of some sort gets started) the child process (0 or leave empty to disable) (higher number = higher stability but also longer delay until the boot loop detection kicks in - if you're unsure or it's unstable, take the biggest number of your measurements and/or add a few seconds)
alwaysKeepAlive: false // set to true to force Node-Wrap to insistently keep alive / restart the child process as fast and reliably as possible (unaffected by boot loop detection though)
}, (time)=>{
// executed on startup of the child process - "time" is the time it took to start the child process
}, (time)=>{
// executed on crash of the child process (status 1) - "time" is the time it took from the initialization of the child process until the crash
}, (time)=>{
// executed on stop of the child process (status 3 or nodeWrap.stop()) - "time" is the time it took from the initialization of the child process until it was stopped
});Even though this package does its job more than perfectly according to my tests (including multiple days long stresstests), I can't guarantee anything.
I try my best to make it as reliable as possible but if you have any complaints or want to suggest a feature, please create a new issue here and I will get to you ASAP.
Additionally, I hereby (and by the provided MIT license) free myself of any liability and legal consequences.
This package was not made with the intention of being malicious so if it is used in criminal activities, I can't be held responsible.
If you modify or share this package you are free to do that under one single condition: It has to still include the "LICENSE" file (also I'd be happy if you contacted me, maybe I'll even "advertise" your version)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent