How to run Cron Job in Node.js application that uses cluster module?

After some research I ended up with "Distributed locks using Redis" solution. There is node module for that: node-redis-warlock.

Hope this answer will be useful for someone else.

UPDATE. Minimal sample code:

var Warlock = require('node-redis-warlock'),
    redis = require('redis');

// Establish a redis client
redis = redis.createClient();

// and pass it to warlock
var warlock = new Warlock(redis);

function executeOnce (key, callback) {
    warlock.lock(key, 20000, function(err, unlock){
        if (err) {
            // Something went wrong and we weren't able to set a lock
            return;
        }

        if (typeof unlock === 'function') {
            setTimeout(function() {
                callback(unlock);
            }, 1000);
        }
    });
}

// Executes call back only once
executeOnce('every-three-hours-lock', function(unlock) {
    // Do here any stuff that should be done only once...            
    unlock();          
});

UPDATE 2. More detailed example:

const CronJob = require('cron').CronJob;
const Warlock = require('node-redis-warlock');
const redis = require('redis').createClient();
const warlock = new Warlock(redis);
const async = require('async');

function executeOnce (key, callback) {
    warlock.lock(key, 20000, function(err, unlock) {
        if (err) {
            // Something went wrong and we weren't able to set a lock
            return;
        }

        if (typeof unlock === 'function') {
            setTimeout(function() {
                callback(unlock);
            }, 1000);
        }
    });
}

function everyMinuteJobTasks (unlock) {
    async.parallel([
        sendEmailNotifications,
        updateSomething,
        // etc...
    ],
    (err) => {
        if (err) {
            logger.error(err);
        }

        unlock();
    });
}

let everyMinuteJob = new CronJob({
    cronTime: '*/1 * * * *',
    onTick: function () {
        executeOnce('every-minute-lock', everyMinuteJobTasks);
    },
    start: true,
    runOnInit: true
});

/* Actual tasks */
let sendEmailNotifications = function(done) {
    // Do stuff here
    // Call done() when finished or call done(err) if error occurred
}

let updateSomething = function(done) {
    // Do stuff here
    // Call done() when finished or call done(err) if error occurred
}

// etc...

If are using PM2, You can use an environment variable provided by PM2 itself called NODE_APP_INSTANCE which requires PM2 2.5 or greater.

NODE_APP_INSTANCE environment variable can be used to determine difference between process, for example you may want to run a cronjob only on one process, you can just do this

if(process.env.NODE_APP_INSTANCE == 0) { //schedule your cron job here since this part will be executed for only one cluster } ,

Since two processes can never have the same number.

More Info on PM2 official doc here.

I actually do not like the redis approach that is also used in the cron-cluster npm plugin, because I do not want to have that redis server running on my maschine and maintain it, too.

I would like to discuss this approach with you:

Pro: we do not need to use redis Con: cron jobs are always running on the same worker

I use the message passing only for this, if you use it for other things, you want to pass the information that

if (cluster.isMaster) {
    // Count the machine's CPUs
    var cpuCount = require('os').cpus().length;;

    // Create a worker for each CPU
    for (var i = 0; i < cpuCount; i += 1) {
        cluster.fork();
    }

    cluster.on('fork', (worker) => {
        console.log("cluster forking new worker", worker.id);
    });

    // have a mainWorker that does the cron jobs.
    var mainWorkerId = null;

    cluster.on('listening', (worker, address) => {
        console.log("cluster listening new worker", worker.id);
        if(null === mainWorkerId) {
            console.log("Making worker " + worker.id + " to main worker");
            mainWorkerId = worker.id;
        worker.send({order: "startCron"});
        }
    });

    // Listen for dying workers if the mainWorker dies, make a new mainWorker
    cluster.on('exit', function (worker, code, signal) {
        console.log('Worker %d died :(', worker.id);

        if(worker.id === mainWorkerId) {
            console.log("Main Worker is dead...");
            mainWorkerId = null;
        }

        console.trace("I am here");
        console.log(worker);
        console.log(code);
        console.log(signal);
        cluster.fork();

    });
// Code to run if we're in a worker process
} else {

    // other code like setup app and stuff

    var doCron = function() {
        // setup cron jobs...
    }

    // Receive messages from the master process.
    process.on('message', function(msg) {
        console.log('Worker ' + process.pid + ' received message from master.', message);
        if(message.order == "startCron") {
            doCron();
        }
    });
}