Node http.createServer how to buffer incoming requests

Question

I'm building a small node server that generates PDF files (using Nightmare.js). Each request calls createPage to generate one pdf.

The incoming request tend to all come around the same time, overloading the PC this is running on.

I need to buffer the incoming requests to delay execution of some requests till some of the current requests have completed. How do I do this?

function createPage(o, final) {
    //generate pdf files
}

http.createServer(function (request, response) {
    var body = [];
    request.on('data', function (chunk) { 
        body.push(chunk);
    }).on('end', function () {
        body = Buffer.concat(body).toString();
        var json = JSON.parse(body);

        createPage(json, function (status) {
            if (status === true) {
                response.writeHead(200, { 'Content-Length': 0 });
                console.log('status good');
            } else {
                response.writeHead(500, { 'Content-Type': 'text/html' });
                response.write(' ' + status);
            }
            response.end('\nEnd of Request \n');
       });
   });
}).listen(8007);

Answer 1

If I understand correctly, you want to continually accept http requests but throttle the rate at which createPage is invoked. If so, you probably need to consider a slightly different design. In this current design, every next client will have to wait longer than the previous one to find out if their request has succeeded or failed.

Approach 1: use a queue (rabbitmq, aws sqs, zeromq, kafka, etc). Here's the basic workflow:

1. receive the request
2. generate a unique id
3. put a message on the queue that includes the data and the unique id
4. return the unique id to the client
5. the client periodically checks for the completion of the task using the unique id

---

Approach 2: Use a queue with message duplexing.

1. receive the request
2. generate a correlation id and relate it to the http transaction
3. send message on queue to worker with correlation id
4. when worker completes, it sends the response back with the correlation id
5. server uses correlation id to find the http transaction and send the appropriate response to the client