restart handling and crash detection

app.js

@@ -17,7 +17,6 @@
 	along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
-
 "use strict";
 /*global require, global, process*/
 

loader.js

@@ -3,6 +3,7 @@
 var	nconf = require('nconf'),
 	fs = require('fs'),
 	cluster = require('cluster'),
+	async = require('async'),
 	numCPUs = require('os').cpus().length;
 
 	/* TODO
@@ -93,36 +94,66 @@ var	nconf = require('nconf'),
 	// },
 	// nbb, nbbOld;
 
+var Loader = {
+	timesStarted: 0
+};
+
+Loader.init = function() {
 	nconf.argv();
 
 	cluster.setupMaster({
 		exec: "app.js",
-	silent: false
+		silent: true
 	});
 
 	for(var x=0;x<numCPUs;x++) {
-	cluster.fork();
+		// Only the first worker sets up templates/sounds/jobs/etc
+		cluster.fork({ cluster_setup: x === 0 });
 	}
 
-cluster.on('exit', function(worker, code, signal) {
-	console.log('worker ' + worker.process.pid + ' died');
-});
-
-Object.keys(cluster.workers).forEach(function(id) {
-	cluster.workers[id].on('message', function(message) {
+	cluster.on('fork', function(worker) {
+		worker.on('message', function(message) {
 			if (message && typeof message === 'object' && message.action) {
 				switch (message.action) {
 					case 'ready':
-					cluster.workers[id].send('bind');
+						console.log('[cluster] Child Process (' + worker.process.pid + ') listening for connections.');
+						worker.send('bind');
+					break;
+					case 'restart':
+						console.log('[cluster] Restarting...');
+						Loader.restart(function(err) {
+							console.log('[cluster] Restarting...');
+						});
 					break;
-				// case 'restart':
-				// 	nbb_restart();
-				// break;
 				}
 			}
 		});
 	});
 
+	cluster.on('exit', function(worker, code, signal) {
+		if (code !== 0) {
+			if (Loader.timesStarted < numCPUs*3) {
+				Loader.timesStarted++;
+			} else {
+				console.log(numCPUs*3, 'restarts in 10 seconds, most likely an error on startup. Halting.');
+				process.exit();
+			}
+		}
+
+		console.log('[cluster] Child Process (' + worker.process.pid + ') has exited (code: ' + code + ')');
+		cluster.fork();
+	});
+};
+
+Loader.restart = function(callback) {
+	async.eachSeries(Object.keys(cluster.workers), function(id, next) {
+		cluster.workers[id].kill();
+		next();
+	}, callback);
+}
+
+Loader.init();
+
 // Start the daemon!
 // if (nconf.get('daemon') !== false) {
 // 	// Check for a still-active NodeBB process

src/middleware/index.js

@@ -129,7 +129,7 @@ module.exports = function(app, data) {
 	routeCurrentTheme(app, data.currentThemeId, data.themesData);
 	routeThemeScreenshots(app, data.themesData);
 
-	if (cluster.worker.id === 1) {
+	if (process.env.cluster_setup === 'true') {
 		meta.templates.compile();
 	} else {
 		setTimeout(function() {

src/routes/debug.js

@@ -56,6 +56,10 @@ module.exports = function(app, middleware, controllers) {
 	});
 
 	router.get('/test', function(req, res) {
-		res.redirect(404);
+		var cluster = require('cluster');
+		console.log('answered by worker', cluster.worker.id);
+		res.send(200);
+		process.exit();
+		// res.redirect(404);
 	});
 };

src/webserver.js

@@ -41,7 +41,7 @@ if(nconf.get('ssl')) {
 	emailer.registerApp(app);
 	notifications.init();
 
-	if (cluster.worker.id === 1) {
+	if (process.env.cluster_setup === 'true') {
 		user.startJobs();
 	}
 
@@ -50,7 +50,7 @@ if(nconf.get('ssl')) {
 		meta.js.minify(app.enabled('minification'));
 		meta.css.minify();
 
-		if (cluster.worker.id === 1) {
+		if (process.env.cluster_setup === 'true') {
 			meta.sounds.init();
 		}
 	});