Client library for autostart-server, automatically exits process on inactivity.


1.2.14 years ago4 years agoMinified + gzip package size for autostart-client in KB



This library isintended as a client for autostart-server. It will automatically shutdown the node process after a defined period of inactivity.

Autostart-Server is inspired by socket-activation via node-systemd, being a pure node implementation of the concept, which is useful for systems where systemd is not an option. It will create net.Servers listening to the the specified sockets and start the corresponding server apps on activity, forwarding all traffic to that app.


This library requires node v.7.6.0 or higher. In principle it should work with node versions since node v.6.5.0, however the tests use newer features, so use at your own risk.

$ npm install autostart-client


The following example will create a client with a timeout of 15 minutes and attach a http server to watch for inactivity

const autostart = require("node-autostart-client");
const server    = require("http").createServer();
const client    = autostart.client({timeout: 15 * 60});

Features are grouped into the basic auto-shutdown functionality and the extended client features.


An AutoShutdown instance contains a timeout, a list of checks and a list of cleanups. After every interval of timeout seconds it will execute registered check. If all checks return truthy values, it will execute all cleanup actions in reverse order. checks and cleanups can be grouped into attachments for easier handling

An AutoShutdown is created using the autostart.autoShutdown(opts) function. opts is an object with the following members:

timeout     Timeout in seconds
checks      Iterable used to prefill checks   (default: [])
cleanups    Iterable used to prefill cleanups (default: []) 
attchments  Iterable containing attachments   (default: [])

In addition to the passed checks and cleanups autostart.exit will be added to cleanups and client.activityCheck is added to checks.

The following will create an AutoShutdown with a timeout of 1000s:

const as = autostart.autoShutdown({timeout: 1000});

Unless indicated otherwise instance methods of an AutoShutdown return the same instance for chaining.

Manually adding and removing checks and cleanups

The addCheck(check), removeCheck(check), addCleanup(cleanup) and removeCleanup(cleanup) methods of an AutoShutdown Instance can be used to add respectively remove checks respectively cleanups. checks and cleanups are all called with zero arguments. checks must return either a truthy value to indicate that the shutdown may proceed or a falsy value to stop it. Both checks and cleanups may return Promises in which case the execution will wait on the promise before continuing with further checks/cleanups and in the case of a check evaluate the resolve-value of the Promise to find out whether to proceed or not.

as.addCheck(() => users.count === 0);
as.addCleanup(async () => await db.close());//close database connection


For typical operation there is often an object, e.g. a server or a database, which compels both checks and cleanups. For easier handling an AutoShutdown allows to group these into so called attachments. An attachment consists of an attachment-object, typically something like a server or database-connection, which is simply used as a key to identify it, as well as a list of associated checks and cleanups. Furthermore it contains a list of actions, which are called in the case of removing an attachment for cleanup purposes. Manually attaching a http-server could for example be done like this:

let lastTime = new Date();
const onReq            = () => lastTime = new Date();
const checkTime        = () => (new Date()).getTime() > lastTime.getTime() + 15 * 60 * 1000;
const checkConnections = () => new Promise((res, rej) => server.getConnections((err, cnt) => res(!err && cnt === 0)));
const clean            = () => new Promise((res, rej) => server.close(() => res()));
server.on('request', onReq);
as.attach(server, [checkTime, checkConnections], [clean], [() => server.removeListener("request", onReq)]);

Calling as.detach(server) afterwards would remove everything added with the attachment from as, as well as executing the action to remove the request listener from server.

Custom Attachments

The autostart.Shutdown.registerAttachmentType(name, func) can be used to attach prebuild attachment functions. It will add func under the name "attach"+name to the prototype of AutoShutDown. An attachment function for websockets via ws could e.g. look kile the following:

autostart.Shutdown.registerAttachmentType("WebSocket", function attachWebSocket(server) {
   let connections = 0;
   const closeConnection = () => { this.resetTimer(); --connections; };
   const onConnect = conn => {
       conn.on("close", closeConnection);
   const check = () => connections === 0;
   const clean = () => new Promise((res, rej) => server.close(() => res()));
   server.on('connection', onConnect);
   return this.attach(server, [check], [clean], [() => server.removeListener('connection', onConnect)]);

It would be called as:

const ws = new WebSocket.Server({});
Prebuild attachments

For ease of use the common use case of attaching a node http-server has been bundled into the as.attachServer(server) method, which behaves similar to the code example, although it uses the connect event in order to support net.Server too.

Further predefined attachments can be activated by requiring files from the attach subdirectory. Currently only the ws.js attachment is bundled, which registers an attachment named Websocket for websocket servers created using ws.js.


const ws = new WebSocket.Server({});

Manual controls

For more finegrained control an AutoShutdown can be controlled manually. The as.stop() method will stop the timer so that no activity checks will be done anymore. as.start() can be used to restart the timer with and therefore restore control to the AutoShutdown. Be advised that as.start() will not continue the previous interval, but start with the full timeout. as.shutdown() can be used to manually trigger the cleanup independent of the current state of the timer.


Each AutoShutdown automatically adds autostart.exit to it's cleanups on construction. autostart.exit will call process.exit(0) to exit the current program. It can be removed by calling as.removeCleanup(autostart.exit).

To simplify building checks there is also a builtin activity measure, which automatically check. This check can be removed using as.removeCheck(as.activityCheck). It queries an internal value, which can be set by calling as.resetTimer(). Checking the time of the last connection from the server-attachment example could therefore be simplified to:

server.on('request', () => as.resetTimer());


An AutostartClient extends the functionality of an AutoShutdown with additional functionality to communicate with an autostart-server. When started it will check, whether the application is launched from an autostart-server, deactivating shutdown functionality using it's own .stop() method otherwise. This makes it easy to execute the script as a standalone in development, but seamlessly put it into an autostart-server for deployment.

An AutoStartClient is created using the autostart.client(options) function. options is the same as when creating an AutoShutdown but also has the following property:

deferInit   If false finishInitialization() will be called during construction (default: false)

An AutoStartClient has all the methods of an AutoShutdown. However unlike a plain AutoShutdown the timeout will not necessarily start on construction. Instead it will wait until both the #asc-init event has been sent from the autostart-server and client.finishInitialization() has been called. Besides starting the timeout client.finishInitialization() will send the #asc-ready event to the autostart-server, signalling that the process is ready to receive connections. This means that client.finishInitialization() should only be called when the app is fully started and listens to its socket.

AutoStartClients have the client.isClient property, which is a boolean indicating if the instance has been started as a client (via nodes child_process) or not.


AutoStartClient has the additional members client.socket, client.connections and All three are promises, which resolve to the data send from the server by the #asc-init event.

  • client.connections resolves to the array of all connections to the app as objects of .src and .dst
  • client.socket resolves to the main socket to which the autostart-server will forward its connections. This is equivalent to (await client.connections)[0].dst
  • contains further custom data coming from the app description.

If you find any bugs or have a feature request, please open an issue on github!

The npm package download data comes from npm's download counts api and package details come from