node package manager

locker-server

locker - lock server written in node.js

This is lock server for system that need to share and lock common resources across servers with sub-second precision.

Features

  • Lock timeouts with millisecond precision:
    • Timeout to wait for getting lock.
    • Timeout to keep lock before releae.
  • No polling: one request to acquire, one request to release.
  • Auto-releasing locks on disconnect.
  • Pure node.js. Just awesome.

Clients

node.js

Client for node.js is completely async.

Example:

var Locker = require("locker"),
    locker = new Locker(4545, "127.0.0.1");
 
locker.on("reset", function() {
    console.log("Reset happened (is server running?)");
});
 
locker.on("error", function(error) {
    console.log("Catched error:", error);
});
 
//            name    wait  max   callback 
locker.locked("five", 2000, 3000, function(error, callback) {
    if (error) {
        // lock failed 
        callback(error);
        return;
    }
 
    // do whatever you want with your shared resource 
    
    callback(undefined, {well: "done"});
});

PHP

Client for php.

Example:

require_once('Locker.php');
 
$Locker = new \Locker\Locker("127.0.0.1", 4545);
 
// Lock creation
$LockOne = $Locker->createLock('example');
 
// getting lock
$LockOne->acquire(200, 10000);
// doing very important stuff
echo 'Waiting for 5 seconds..'."\n";
sleep(5);
// releasing lock
$LockOne->release();

Running

First create a dir for locker:

mkdir locker
cd locker
mkdir node_modules
npm install locker-server

Then create main server file (look at example/server.js) and save it as server.js:

(function() {
    var Locker = require("locker-server"),
        locker = new Locker();
 
    locker.listen(4545);
})();

Run it after all:

node server.js

Capacity

More performance tests to be done, but for now on my MacBook Air (Intel i5):

Clients count: 5
Different locks to request: 10000
Total locks to request: 50000
Lock work time: 1ms
 
Server CPU usage: 100%
Server RSS memory: 70-80M (no growth after 5M requests)
Responses per second: 15000-17000 (measuring every 5 seconds)

If you want more, shard your locks.