BufferedReader.read() eating 100% of CPU

Question

I have a JAVA game server that uses 1 thread per TCP connection. (I know it's bad but i'll have to keep it this way for now). On a (3.2Ghz 6cor x2 machine, 24GB RAM, windows server 2003 64bits) and here is a piece of the code:

public void run()
{
    try
    {   
        String packet = "";
        char charCur[] = new char[1];

        while(_in.read(charCur, 0, 1)!=-1 && Server.isRunning)
        {
            if (charCur[0] != '\u0000' && charCur[0] != '\n' && charCur[0] != '\r')
            {
                packet += charCur[0];
            }else if(!packet.isEmpty())
            {
                parsePlayerPacket(packet);
                packet = "";
            }
        }

    }catch(Exception e)
    {
        e.printStackTrace();
    }
    finally
    {
        try{
            kickPlayer();
        }catch(Exception e){e.printStackTrace();};

        Server.removeIp(_ip);
    }
}

After about 12 hours or more of server upTime (and about 3.000 players connected) the server starts eating 100% of all the 12 CPUs for ever, until I manually reboot the JAVA application. So the game starts lagging verry bad and my players starts complaining.

I have tried profiling the application and here is what I came up with:

So I am guessing that the problem is coming from here:

while(_in.read(charCur, 0, 1)!=-1 && Server.isRunning)

knowing that the variable "_in" is a reader of the socket input : (_in = new BufferedReader(new InputStreamReader(_socket.getInputStream()))).

Why on earth _in.read() takes so much CPU after a long server upTime?

I have tried putting a Thread.sleep(1); and more inside the While loop, but doesn't do anything, I guess the problem is inside of the BufferedReader.read() method.

Does anyone have any idea of what can cause this?? And how to fix it?

Answer 1

This is a duplicate of your previous question: An infinite loop somewhere in my code. Please do not open up a new question, but instead use the editing functions.

That being said, 3000 threads is definitely a lot and would most likely cause excessive amounts of context switching. Instead of starting a new thread for each connection, consider using non-blocking IO facilities in Java. Examples can be found here: http://download.oracle.com/javase/1.4.2/docs/guide/nio/example/index.html