问题
I have a simple code for a multi-threaded echo server in Java (it returns whatever received back to the clients). I'm profiling various resources of the server including the thread statistics. Below are some of these statistics as per number of connected clients. My questions is for the baseline (# of clients 0) compared with non-baselines!
1) why when a single client connects, the total thread count increases by 2? (for the rest, it makes sense to increment by 1)
2) What are the two non-daemon threads?! And why the daemon initially increment by 1 and then is fixed?
Are are they kind of random?!
# clients 0 1 2 3 4 5 6 7 8 9 10
Total Started Thread Count 15 18 19 20 21 22 23 24 25 26 27
Thread count 14 16 17 18 19 20 21 22 23 24 25
Peak thread count 14 16 17 18 19 20 21 22 23 24 25
Daemon thread count 12 13 13 13 13 13 13 13 13 13 13
Here is the piece of code for the server. I'm using both RMI (for clients to poll messages) and Server Socket (for clients to send messages). If other classes are needed let me know.
package test;
import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStreamReader;
import java.io.PrintWriter;
import java.net.ServerSocket;
import java.net.Socket;
import java.rmi.registry.LocateRegistry;
import java.rmi.registry.Registry;
import java.rmi.server.UnicastRemoteObject;
import java.util.Vector;
public class ServerRMI extends Thread implements Hello {
//centralized token manager runs polling server and socket server to receive updated tokens
static Vector<String> tokenList= new Vector<String>();
protected Socket clientSocket;
static int RMIRegistryPort=9001;
static int SocketServerPort=9010;
public static void main(String[] args) throws IOException {
try {
ServerRMI obj = new ServerRMI();
Hello stub = (Hello) UnicastRemoteObject.exportObject(obj, 0);
// Bind the remote object's stub in the registry
Registry registry = LocateRegistry.createRegistry(RMIRegistryPort);
registry.bind("Hello", stub);
System.err.println("Server ready");
} catch (Exception e) {
System.err.println("Server exception: " + e.toString());
e.printStackTrace();
}
ServerSocket serverSocket = null;
//initialize token list
//A needs to execute first
tokenList.add(0,"0");
try {
serverSocket = new ServerSocket(SocketServerPort);
System.out.println("Connection Socket Created");
try {
while (true) {
System.out.println("Waiting for Connection");
new ServerRMI(serverSocket.accept());
}
} catch (IOException e) {
System.err.println("Accept failed.");
}
} catch (IOException e) {
System.err.println("Could not listen on port: "+SocketServerPort);
} finally {
try {
serverSocket.close();
} catch (IOException e) {
System.err.println("Could not close port: "+SocketServerPort);
}
}
}
private ServerRMI(Socket clientSoc) {
clientSocket = clientSoc;
start();
}
public ServerRMI() {}{
// TODO Auto-generated constructor stub
}
public void run() {
System.out.println("New Communication Thread Started");
try {
PrintWriter out = new PrintWriter(clientSocket.getOutputStream(),
true);
BufferedReader in = new BufferedReader(new InputStreamReader(
clientSocket.getInputStream()));
String inputLine;
while ((inputLine = in.readLine()) != null) {
tokenList.add(0,inputLine);
System.out.println("Server received: " + inputLine);
// System.out.println(" ququ size: "+queue.size());
out.println(inputLine);
if (inputLine.equals("Bye."))
break;
}
out.close();
in.close();
clientSocket.close();
} catch (IOException e) {
System.err.println("Problem with Communication Server");
}
}
public String pollServer() {
if(!tokenList.isEmpty()){
String data = tokenList.get(0);
System.out.println("Poll data: "+data);
return data;
} else{
return tokenList.size()+"";
}
}
}
回答1:
I'm not sure what you use to receive those connections, but usually frameworks that handle TCP connections non-blocking like Netty use a master thread to listen to the port and a thread pool to handle incomming connections. This means it will at least open 2 threads for the incomming connection if the thread-pool is limited to 1 thread.
See this example from the netty page, where 2 NioEventLoopGroups are used while bootstrapping the server.
The 2 threads are necessary to not block incoming traffic.
回答2:
I just hope you reviewed the Thread names in VisualVM profiler. Usually the thread names give you some idea of what is started.
ad 1) this could very well be some TCP/IP background (cleaner, poller) thread that gets spawned as soon as you have some TCP/IP connection from outside.
What does happen when clients go back to 0. Does extra thread disappear?
Answers may very well differ based on JVM vendor and version. So I think you need to look in JVM internals about sockets and threads.
Example below is more straight forward (without RMI)
import java.net.*; // for Socket, ServerSocket, and InetAddress
import java.io.*; // for IOException and Input/0utputStream
public class TCPEchoServer {
private static final int BUFSIZE = 32; // Size of receive buffer
public static void main(String[] args) throws lOException {
if (args.length != i) // Test for correct # of args
throw new lllegalArgumentException("Parameter(s): <Port>");
int servPort = Integer.parselnt(args[0]);
// Create a server socket to accept client connection requests
ServerSocket servSock = new ServerSocket(servPort);
int recvMsgSize; // Size of received message
byte[] byteBuffer = new byte[BUFSlZE]; // Receive buffer
}
for (;;) { // Run forever, accepting and servicing connections
Socket clntSock = servSock.accept(); // Get client connection
System.out.println("Handling client at " +
clntSock.getInetAddress().getHostAddress() + " on port " +
clntSock, getPort());
InputStream in = clntSock, getlnputStream();
OutputStream out = clntSock.getOutputStream();
// Receive until client closes connection, indicated by-i return
while ((recvMsgSize = in .read(byteBuffer)) != -I)
out.write(byteBuffer, O, recvMsgSize);
clntSock, close();
}
}
来源：https://stackoverflow.com/questions/37154657/daemon-threads-thread-count-and-total-started-thread-count