MCNodeSrv - Dashboard

Presenting: MCNodeSrv

A minecraft server wrapper written in NodeJS, using the express framework.

I wrote it as a replacement for giving away SSH access, which can screw a box up alot harder if done right/wrong.

Features thus far include:

• Webinterface (derp) styled by Bootstrap
• Starting/Stopping/Killing the server process
• Web-based settings for changing command line arguments as well as environment variables of the server process
• Monitor stdout/stderr (it even highlights lines with the word ‘error’ in it, cool eh?)
• (planned) Send data to stdin
• JSON based config file(s)

… and I’m not done yet.
Next up will be hardening the node.js instance so it won’t take the minecraft server down with it, if it should happen to crash. I’m also thinking about adding a command-line interface to do simple stuff.. oh! *mental note: gracefully stop server process on ^C and other harmful signals*

Bottom line: NodeJS is just awesome.

var a = {"test": "moo"}; // Define an object
var b = a; // Copy the object
b.test = "meh";
alert(a.test); // alert: "meh"
// wait- what? b = a makes a reference instead of a copy, doh!

One way to get around this is:

var b = JSON.parse(JSON.stringify(a)) // This is one ugly hack

but it feels.. wrong.. and probably doesn’t work with nested functions and stuff like that.

But I just made a groundbreaking discovery. The reason I’ve always been getting “no supported authentication methods left to try is” *drum roll* :

I added my private key for work to the putty “Default Settings” session, which (for some reason) prevents Putty (and in turn PLink) from falling through to Pageant and fails when it tries to authenticate to github with my work-SSH-key.

God damn! I’ve been working on getting TortoiseGit to work for so long, been trying different versions of each component and in the end it was as simple as clearing a textbox and clicking a checkbox.

I noticed that the Aprilon MineCraftApproval system got slower and slower over time so I whipped up a crude profiling mechanism, extending the mysqli class with microtime() time-tracking code and function call-frequency-tracking code using debug_backtrace().
It also prints out  every MySQL query made, the call-stack which led to this query and sorts everything by how long it took for the each transaction.
Bear in mind that this includes processing overhead by the MySQLi PHP extension, libmysql library on the system and sockets opening/closing, receiving and sending if the database is on a remote server.

You can see example output on the right and you may notice a few warnings regarding “No possible_keys”.
That means the query could not use any indices because the table doesn’t have any matching the WHERE clause (cut off on pic.).

The number of warnings shrunk to zero after a few ALTER TABLE, ADD INDEX(..),
and the time to execute the MySQL queries was reduced from 59ms (wolfram: ~blink of an eye) to 16.17ms (wolfram: ~nerve impulse to travel the length of a human).

Excuse my french, but that’s fuckin’ impressive.

Even if’s ‘just’ a difference of  42.84ms, it still makes an impact on the experience of the whole application.

Since theres no official documents on API access (or they are well hidden) I had to decompile the Java plugin they provide and reverse engineer the protocol.

Half an hour later I came up with this:

Its spooky how much fun collaborating projects with mercurial can be, even though its just me and Karl (Aprilon.net).