Error Handling in Node.js

Paul A. Jungwirth

NodePDX

May 2013

Laugh Track Server

• Asynchronous
• Webhook callback to indicate success or failure

No error handling

function encodeLaughTracks(infile) {
  for (var i = 1; i <= 3; i++) {
    var cmd = "../laugh.rb -q " + i + " " + infile;
    console.log(cmd);
    exec(cmd, function (error, stdout, stderr) {
      if (error) throw error;
      notifyCaller(infile, true);
    });
  }
}   

function notifyCaller(infile, success) {
  console.log("finished adding laughs to " + infile);
  // HTTP POST to indicate success or failure
}

• Lost requests
• No stack context

Callbacks vs. Stack

/home/paul/src/node-errors-talk/src/orig/app.js:10
    if (error) throw error;
                     ^
Error: Command failed: 
    at ChildProcess.exithandler (child_process.js:634:15)
    at ChildProcess.EventEmitter.emit (events.js:98:17)
    at maybeClose (child_process.js:732:16)
    at Process.ChildProcess._handle.onexit (child_process.js:799:5)

forever.js?

• Automatically restarts node whenever it dies.
• That exception will still kill our server, so now the jobs will never finish.
• Restart the server to keep handling requests, but all current requests just died.

listen for uncaughtException?

process.on('uncaughtException', function(err) {
  console.log(err);
});

• Which request threw it?
• What is our state?
• Deprecated.

try catch?

try {
  exec(cmd, function (error, stdout, stderr) {
    if (error) throw error;
    notifyCaller(infile, true);
  });
} catch (e) {
  console.log(e);
  notifyCaller(infile, false);
}

• But nothing outside the callback ever sees the exception.
• Callbacks start a new stack, so there is nothing to unwind!

Domains

• Since Node 0.8
• Unstable
• Error callbacks are scoped (sort of), not global.

Domains

var domain = require('domain');

http.createServer(function(request, response) {
  var d = domain.create();
  var queryData = url.parse(request.url, true).query;
  var infile = queryData['file'];

  d.on('error', function(err) {
    console.log(err);
    notifyCaller(infile, false);
  });

  d.run(function() {
    /* ... */
  });
}).listen(8086);

• Still can leak memory
• Need to kill the ps on error
• Still no stacktrace

Cluster + Domains

• Separate ps for each Cluster.
• Kill just the ps that error'ed.
• Give other requests time to finish first.
• One ps per core?
• Experimental

Just Cluster

var cluster = require('cluster');
var cpus = require('os').cpus().length;

if (cluster.isMaster) {
  // fork workers
  for (var i = 0; i < cpus; i++) {
    cluster.fork();
  }

  cluster.on('exit', function(worker, code, signal) {
    console.log('worker ' + worker.process.pid + ' died');
  });

} else {
  http.createServer(function(request, response) {
    /* ... */
  }).listen(8086);
}

Cluster + Domains

if (cluster.isMaster) {
  for (var i = 0; i < cpus; i++) {
    cluster.fork();
  }

  cluster.on('disconnect', function(worker) {
    console.log('worker ' + worker.process.pid + ' died');
    cluster.fork();
  }); 

} else {
  http.createServer(function(request, response) {
    var d = domain.create();
    var infile = /* ... */;
    d.on('error', function(err) {
      console.log(err.stack);
      notifyCaller(infile, false);

      try {
        // Ten minutes to let other connections finish:
        var killTimer = setTimeout(function() {
          process.exit(1);
        }, 10 * 60000);
        killTimer.unref(); // Don't stay up just for the timer
        cluster.worker.disconnect(); // Stop taking new requests
      } catch (err2) {
        console.log("Error handling error!: " + err2);
      } 
    });

    d.add(request);  // Explicit binding
    d.add(response); // Explicit binding

    d.run(function() {
      /* ... */
    });
  }).listen(8086);
} 

Full Stack Traces with longjohn

// package.json:
{  
  "name": "NodeExceptionsDemo",
  "dependencies": { 
    "longjohn": ""
  } 
}   

// app.js:
require('longjohn')

longjohn Stack Trace:

Error: Command failed: 
    at exithandler (child_process.js:634:15)
    at EventEmitter.emit (events.js:98:17)
    at maybeClose (child_process.js:732:16)
    at Socket.<anonymous> (child_process.js:945:11)
    at EventEmitter.emit (events.js:95:17)
    at Pipe.close (net.js:451:12)
---------------------------------------------
    at exports.execFile (child_process.js:688:9)
    at exports.exec (child_process.js:575:18)
    at encodeLaughTracks (/home/paul/src/node-errors-talk/src/longjohn/app.js:14:5)
    at /home/paul/src/node-errors-talk/src/longjohn/app.js:56:9
    at b (domain.js:183:18)
    at Domain.run (domain.js:123:23)
    at Server.<anonymous> (/home/paul/src/node-errors-talk/src/longjohn/app.js:52:7)

References and Further Reading

• http://nodejs.org/api/domain.html
• http://nodejs.org/api/cluster.html
• http://shapeshed.com/uncaught-exceptions-in-node/
• http://stackoverflow.com/questions/7310521/node-js-best-practice-exception-handling
• https://groups.google.com/forum/?fromgroups#!msg/nodejs/5R6kPncBGwE/woBz84kwMcAJ
• https://github.com/mattinsler/longjohn
• https://github.com/pjungwir/node-errors-talk