问题
Is there a method to know the port number of both the client and server during RMI?
When a result is returned during RMI to the client, next time the client requests the result, will the conversation between client and server be at the same port as the previous, when the server has been running since the first result was returned or a new port is created?
for example :
// call to a remote method add
addServerIntf.add(d1,d2)
// after this call i get the added result
// again call the add method by again starting the client
addServerIntf.add(d1,d2)
// I get the result back as usual
In the 2 different calls does the port number of client and server remain the same ?
My client program exits after entering a command like java AddClient localhost 100 200 The method on the server returns 300 and the client exits. Next time i start my client again with java AddClient localhost 19 100. Now will request be sent from the same port as sent from before and will the server receive the request on the same port ? Or the situation is different from what i just wrote ?
回答1:
Is there a method to know the port number of both the client and server?
No. You can't know the client port number in advance, as it is allocated dynamically on connection, and the server IP:port is embedded in the stub where you can't get at it. Why would you need to know? The information can't do you much good: you can't use it.
When a result is returned during RMI to the client, next time the client requests the result, will the conversation between client and server be at the same port as the previous, when the server has been running since the first result was returned or a new port is created?
Either:
- The client connection is reused, with the same port number, or
- A new connection with probably a new local port is created at the client.
It is impossible to tell which. The server port remains fixed.
In the 2 different calls does the port number of client and server remain the same ?
See above.
My client program exits after entering a command like
java AddClient localhost 100 200. The method on the server returns 300 and the client exits. Next time i start my client again withjava AddClient localhost 19 100. Now will request be sent from the same port as sent from before
Probably not.
and will the server receive the request on the same port?
Impossible to say without seeing your server code. If you exported it on a constant port, it is exported on a constant port. If not, not. Tautology really.
Or the situation is different from what i just wrote?
I don't understand the question.
回答2:
Is there a method to know the port number of both the client and server during RMI?
If I didn't misunderstand, I think there's no way to know the port number during a RMI session.
If you have a different port number, from the default RMI registry Server 1099 port, you have to set it, on the Server class and on the Client class, because, as the Oracle RMI page reports:
If the registry will be running on a port other than 1099, you'll need to specify the port number in the calls to LocateRegistry.getRegistry in the Server and Client classes. For example, if the registry is running on port 2001 in this example, the call to
getRegistryin the server would be:
Registry registry = LocateRegistry.getRegistry(2001);
and:
This client first obtains the stub for the registry by invoking the static LocateRegistry.getRegistry method with the hostname specified on the command line. If no hostname is specified, then null is used as the hostname indicating that the local host address should be used.
Then, my conclusion is that you can't know, with a method, the port number of a RMI session (you can check the RMI API for details), except when you have to set it, if it is different from the default RMI registry Server 1099 port, because you have to know it at the beginning of the RMI session.
Just think: How you can get this port number? Contacting the server or the client? For example, if you request a page located in a specific Server, that is listening to port 81 (not on the default 80 port), you need in advance the port number to connect to this specific Server, contacting it, for example on: http://192.168.1.1:81. Then, during a RMI session you must know in advance the RMI registry port.
Check this page for details.
When a result is returned during RMI to the client, next time the client requests the result, will the conversation between client and server be at the same port as the previous, when the server has been running since the first result was returned or a new port is created?
When the result is returned to the client, the conversation between Client and Server should be shared on the same RMI registry port, otherwise, if the RMI registry port set in the client is different from the Server RMI registry port (if I didn't forgot), the code will throw a RemoteException, that can occur when a failure occurs in the RMI process.
UPDATE
Now I see your updated question.
In the 2 different calls does the port number of client and server remain the same ?
It should be the same RMI registry Server port. When your program exits after the first call, the program closes the socket connection. The next time you start the program, the RMI registry port should be the same. Otherwise, your program should throw an exception, or, when you pass your arguments to the program, you get an unexpected result. If I understood, your client program simply calls a sum method on the Server. After the first result, the next time you start the program, do you get a different result? If not, I think the RMI registry port is the same.
回答3:
Know your TCP/IP: A client connects to a server on a certain port. The client port number is random (or it should be for security reasons), the server port number is known (otherwise, there would be no way for the client to connect in the first place).
The TCP/IP protocol will then establish a connection. The server will create a handler for the connection and TCP/IP will assign this connection a new, random port. The port number on the server side doesn't change. Since the server can use the client's IP+port as a key to distinguish different "sessions", any number of clients can connect to the same server port.
This means: After the RMI framework is initialized, both client and server will talk to each other over a pair of ports that doesn't change as long as the connection exists.
If the connection is severed for some reason, a new connection can be established but that will probably get a new port numbers on both sides the client side.
my client program exits
This means the connection is severed. When you start the client again, a new client port is created. It most likely (at a chance of 65534:65535 which is roughly 1) will be a new port number.
If you start the client again within two minutes, the port numbers must be different because the old ports are is kept alive for two minutes to make sure the "remote TCP received the acknowledgment of its connection termination request"
(if you run netstat, you will see these with the status TIME_WAIT).
来源：https://stackoverflow.com/questions/8591990/knowing-the-port-number-during-rmi