Skip to content

conpagoaus/exiting

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

61 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Exiting

Safely shutdown hapi.js servers whenever the process exits.

Node.js CI

Details

While it is simple to start and stop a server, ensuring proper shutdown on external, or internal, triggers can be cumbersome to handle properly. exiting makes this easy by managing your Hapi servers, taking care of starting and stopping them as appropriate.

Depending on the exit trigger, the hapi servers will either be gracefully stopped or aborted (by only triggering onPreStop hooks). The exit triggers are handled as detailed:

  • Graceful exit with code 0:
    • process.exit() with exit code 0.
    • Unhandled SIGINT kill signal, through eg. ctrl-c.
    • Unhandled SIGTERM kill signal.
    • Unhandled SIGQUIT kill signal.
  • Aborted exit:
    • process.exit() with non-zero exit code.
    • Unhandled SIGHUP kill signal (code 1).
    • Any uncaught exception (code 1).
    • Any unhandled rejection (code 1).
    • Any closed connection listeners, eg. on worker disconnect (code 255).

If shutting down one of the servers is too slow, a timeout will eventually trigger an exit (exit code 255).

The shutdown logic is programmed to handle almost any conceivable exit condition, and provides 100% test coverage. The only instances that onPreHook code is not called, are uncatchable signals, like SIGKILL, and fatal errors that trigger during shutdown.

Example

Basic server example:

const Hapi = require('hapi');
const Exiting = require('exiting');

const server = Hapi.Server();
const manager = Exiting.createManager(server);

server.events.on('stop', () => {

    console.log('Server stopped.');
});

const provision = async () => {

    server.route({
        method: 'GET',
        path: '/',
        handler: () => 'Hello' 
    });

    await manager.start();

    console.log('Server started at:', server.info.uri);
};

provision();

The server and process life-cycle will now be managed by exiting.

If you need to delay the shutdown for processing, you can install an extention function on the onPreStop or onPostStop extension points, eg:

server.ext('onPreStop', () => {

    return new Promise((resolve) => {

        setTimeout(resolve, 1000);
    });
});

Multiple servers example:

const Hapi = require('hapi');
const Exiting = require('exiting');

const publicServer = Hapi.Server();
const adminServer = Hapi.Server();
const manager = Exiting.createManager([publicServer, adminServer]);

const provision = async () => {

    publicServer.route({
        method: 'GET',
        path: '/',
        handler: () => 'Hello'
    });

    adminServer.route({
        method: 'GET',
        path: '/',
        handler: () => 'Hello Admin'
    });

    await manager.start();

    console.log('Public server started at:', publicServer.info.uri);
    console.log('Admin server started at:', adminServer.info.uri);
};

provision();

Installation

Install using npm: npm install exiting.

Usage

To enable exiting for you server, replace the call to server.start() with Exiting.createManager(server).start().

Exiting.createManager(servers, [options])

Create a new exit manager for one or more hapi.js servers. The options object supports:

  • exitTimeout - When exiting, force process exit after this amount of ms has elapsed. Default: 5000.

await manager.start()

Starts the manager and all the managed servers, as if server.start() is called on each server. If any server fails to start, all will be stopped with server.stop() before the error is re-thrown.

Note that process.exit() is monkey patched to intercept such calls. Starting also installs the signal handlers and an uncaughtException handler.

await manager.stop([options])

Stops the manager and all the servers, as if server.stop() is called on each server.

Notes on process.exit()

The process.exit() method is handled in a special manner that allows the asyncronous stop logic to resolve before actually exiting. Since this can be called from anywhere in the code, and subsequent code is never expected to be executed, the manager will throw an Exiting.ProcessExitError to attempt to escape the current execution context. This allows something like the following to still exit:

while (true) {
    process.exit(1);
}

This might not always work, and can potentially cause a lock up instead of exiting. Eg. with this code:

try {
    process.exit(1);
}
catch (err) {
    /* do nothing */
}
while (true) {}

You should avoid using process.exit() in your own code, and call manager.stop() instead.

Packages

No packages published

Languages

  • JavaScript 97.2%
  • TypeScript 2.8%