pinba

Pinba for Node.js

Node-pinba


Pinba module for node.js.

Pinba is a MySQL storage engine that acts as a realtime monitoring/statistics server using MySQL as a read-only interface. Node-pinba provides client for Pinba server that provides statistics sending from your node.js application.

Check out the Github repo for the source code.

You can install this module via npm:

$> npm install pinba

Node.js is not a bottleneck anymore! Now you can use pinba to measure particular code parts and collect information about requests to your server.

Because Node.js operates in single-threaded event loop and does not orient only for HTTP servers, there was some differences between this module and original Pinba for PHP. Mostly, there is no isolate requests in Node.js, so you should take care of creating Pinba.Request instance for each code chain that you mention as request and call PinbaRequest instance flush method at response end.

There is a simple example how to capture timer value around database call during request:

var http = require('http');
var PinbaRequest = require('pinba').Request;
var db = require('./my_db.js');
 
http.createServer(function (reqres) {
  var pr = new PinbaRequest({
    server_name: 'example.com',
    script_name: '/handler'
  });
 
  var timerDb = pr.timerStart({type: 'db', op: 'select'});
  db.query("SELECT data FROM database", function (errdata) {
    pr.timerStop(timerDb);
 
    if (err) throw err;
 
    res.writeHead(200, {'Content-Type': 'text/plain'});
    res.end('Hello World\n');
    pr.flush();
  });
}).listen(8080, '127.0.0.1');

Read this module API docs and original Pinba docs for more information.

Connect Pinba middleware can be founded here. Feel free to write any for other frameworks.

There was some difference between original PHP threaded model and Node.js event-loop requests processing. So not all original Pinba features are supported.

Fully supported:

    string hostname
    string server_name
    string script_name
    string schema
 
    uint32 request_count
    float request_time
 
    string dictionary
 
    uint32 tag_name
    uint32 tag_value
 
    uint32 timer_hit_count
    float timer_value
    uint32 timer_tag_count
    uint32 timer_tag_name
    uint32 timer_tag_value

Allowed to be passes to flush() method:

    uint32 document_size
    uint32 status

Cannot be implemented properly:

    float ru_utime
    float ru_stime
    uint32 memory_peak
    uint32 memory_footprint

To contribute any patches, simply fork this repository using GitHub and send a pull request to me. Thanks!

MIT license. See license text in file LICENSE.