CODEWITHSUNDEEP
node.jsfs
user393219
user393219

Reputation:

Exit Node Process After Successful fs.appendFile

I'm having trouble create processes in parallel with Node while exiting when they're done with a simple HTTP GET request. I've noticed that if I fire a process.exit() inside of a callback for appendFile, some files will not be created or appended in a Node cluster setup. Ideally, the way below is how I would like to fire events since the process is exited as soon as the job is done:

var rp = require("request-promise");
    config = require("./config"),
    cluster = require("cluster"),
    os = require("os"),
    fs = require("fs");

var keywordArray = [
    'keyword1',
    'keyword2',
    ...
];

if (cluster.isMaster) {

    var numCPUs = os.cpus().length;
    var clusterDivision = Math.ceil(keywordArray.length/numCPUs);

    // Reset the json if previously set
    keywordArray.forEach(function(arrayItem) {
        fs.unlink(config.dataDirectory + arrayItem + '.json', function(err) {
            if (err) console.error(err);
            console.log('successfully unlinked ' + arrayItem + '.json from ' + config.dataDirectory);
        });
    });

    // Create a worker for each CPU
    // Seperate the array out evenly for each worker
    for (var j=1;j<=numCPUs;j++) {
        var tempArray = [];
        var removed = keywordArray.splice(0, clusterDivision);
        if (removed.length > 0) {
            // The array contains something so let's do something with the keyword
            console.log('creating a worker');
            cluster.fork().send(removed);
        } else {
            // We don't need a cluster here
        }
    }

    process.on('exit', function() {
        console.log('exited');
    });

} else if (cluster.isWorker) {
    //  Code to run if we're in a worker process

    // Send the object we created above from variables so they're available to the workers
    process.on('message', function(seperatedArrayItem) {

        seperatedArrayItem.forEach(function(arrayItem) {
            function radarRequest(err, response, body) {
                var responseBody = JSON.parse(body);
                console.log(arrayItem); 
                fs.appendFileSync(config.dataDirectory + arrayItem + '.json', JSON.stringify(responseBody.results, null, '\t'), function (err) {
                    if (err) console.err(err);
                    console.log('success writing file');
                });
            }

            rp({
                url: config.radarSearchURI + 
                '?key='+ config.apiKey + 
                '&location=' + config.latitude + ',' + config.longitude + 
                '&radius=' + config.searchRadius + 
                '&keyword=' + arrayItem, headers: config.headers
            }, radarRequest);
        });

        setTimeout(function() {
            process.exit(0);
        }, 5000);
    });
}

The only way I can make sure all files are properly appended is by using a Timeout, which is exactly what I don't want to - and shouldn't - do. Is there another way I can ensure an appendFile has happened successfully and then kill the node process? Here's a way that works (assuming the process doesn't take longer than 5 seconds):

    process.on('message', function(seperatedArrayItem) {

    seperatedArrayItem.forEach(function(arrayItem) {
        function radarRequest(err, response, body) {
            var responseBody = JSON.parse(body);
            console.log(arrayItem); 
            fs.appendFile(config.dataDirectory + arrayItem + '.json', JSON.stringify(responseBody.results, null, '\t'), function (err) {
                if (err) console.err(err)
                console.log('success writing file');
            });
        }

        rp({
            url: config.radarSearchURI + 
            '?key='+ config.apiKey + 
            '&location=' + config.latitude + ',' + config.longitude + 
            '&radius=' + config.searchRadius + 
            '&keyword=' + arrayItem, headers: config.headers
        }, radarRequest);
    });

    setTimeout(function() {
        process.exit(0);
    }, 5000);
});

Upvotes: 1

Views: 1103

Answers (2)

ranieribt
ranieribt

Reputation: 1306

Node fs.appendFile( ... ) is an asynchronous function. So it expects us to pass a callback for we know it has finished its main operation, to inform us of some error occurred, or another purpose.

This means we need to call Node process.exit( ... ) in the scope of the provided callback. I've written this code to test:

'use strict';

var fs = require('fs');

function jsonValue(obj) {
    return JSON.stringify(obj, null, '\t');
}

fs.appendFile('file.json', jsonValue(['t', 'e', 's', 't']), function(error) {
    if (error) {
        throw error;
    }

    console.log('success writing file');  // no error, so log...
    process.exit();                       // and exit right now
    console.log('exited?');               // this will not be printed
});

Well, it worked as defined.

Other way it works is to use the synchronous version of fs.appendFile( ... ) and call process.exit() in a sequential way:

fs.appendFileSync('file.json', jsonValue(['t', 'e', 's', 't']));

console.log('success writing file'); // no error (I hope so =), so log...
process.exit(); // and exit right now
console.log('exited?'); // this will not be printed

This is clean code and works, but you lose the robustness and convenience gained with the callback...

Upvotes: 2

loganfsmyth
loganfsmyth

Reputation: 161457

You can use an async flow control module like async to kill the process after all files are written. I'd also recomment cluster.worker.disconnect() so that the node process will simple exit gracefully, but that isn't a requirement.

async.forEach(seperatedArrayItem, function(item, done){
    // append file and call 'done' when it is written.

}, function(){
    // Will be called when all item 'done' functions have been called.
    cluster.worker.disconnect();
});

Upvotes: 2

Related Questions