Restart on EADDRINUSE

[tshemsedinov]

Describe the problem

uncaughtException: Error: listen EADDRINUSE: address already in use 0.0.0.0:8000
  Server.setupListenHandle [as _listen2] (node:net:1463:16)

Describe the solution

Port is in use by another server, so we can:

• wait for timeout
• try 3 times
• exit 1

[MarhiievHE]

@tshemsedinov It looks like the HTTP server is not returning this error via the error listener. Will it be ok to catch this error in this way in worker.js, or do we need to find a more elegant solution?:

    process.on('uncaughtException', function(err) {
      if (err.code === 'EADDRINUSE') {
        console.warn(`Address in use: ${options.host}:${port}, retry...`);
        setTimeout(() => {
          tryToStartCounter++;
          if (tryToStartCounter > 3) process.kill(process.pid, 'SIGINT');
          application.server = new metarhia.metacom.Server(application, options);
        },options.timeouts.bind);
      }
    });

https://github.com/MarhiievHE/impress/blob/594c068dc9e97a9c3734dcc44251e32814aa5492/lib/worker.js#L86-L102

[MarhiievHE]

@tshemsedinov It looks like the HTTP server is not returning this error via the error listener. Will it be ok to catch this error in this way in worker.js, or do we need to find a more elegant solution?:

    process.on('uncaughtException', function(err) {
      if (err.code === 'EADDRINUSE') {
        console.warn(`Address in use: ${options.host}:${port}, retry...`);
        setTimeout(() => {
          tryToStartCounter++;
          if (tryToStartCounter > 3) process.kill(process.pid, 'SIGINT');
          application.server = new metarhia.metacom.Server(application, options);
        },options.timeouts.bind);
      }
    });

https://github.com/MarhiievHE/impress/blob/594c068dc9e97a9c3734dcc44251e32814aa5492/lib/worker.js#L86-L102

I was wrong, that wasn't the point. The problem was that a wsServer was created in addition to the httpServer, and the error occurred in the wsServer, so the httpServer error listener did not catch it.

[MarhiievHE]

It makes three tries to open the server on the specified port with bind timeout and then sends SIGTERM, but it doesn`t stop the application with error code 1!

[tshemsedinov]

Fixed in #1908 and published 3.0.7