Programming

Hi,

I am currently working on a Server/Client application on a cluster

I am running about 100 clients in parallel and one server (32 clients per node with 8 processor => 4 clients per processor) . The server is linked to all the clients via Sockets.

After about 55 iteration (about 20 mins), i have a client dying and i receive the following error:

Client 126 java.net.SocketException: Connection reset
at java.net.SocketInputStream.read(SocketInputStream. java:16
at java.net.SocketInputStream.read(SocketInputStream. java:182)
at java.io.ObjectInputStream$PeekInputStream.peek(Obj ectInputStream.java:2249)
at java.io.ObjectInputStream$BlockDataInputStream.pee k(ObjectInputStream.java:2542)
at java.io.ObjectInputStream$BlockDataInputStream.pee kByte(ObjectInputStream.java:2552)
at java.io.ObjectInputStream.readObject0(ObjectInputS tream.java:1297)
at java.io.ObjectInputStream.readObject(ObjectInputSt ream.java:351)
at ServerClient.Server$TcpThread.run(Server.java:789)
Connection reset


the Server.java.789 is : ServerThread.sockIn.readObject();


Do you have any idea what could create the connection reset ? The other clients are still running. It seems that the client socket is creating an exception when we try to read a file which actually does not exist.

My main question is what are the method to identify the problem related to this specific socket (note that I am working on a cluster, hence accessing the node via ssh connection and without having admin access).

I dont understand why suddenly one client would die. It was working well during 20 minutes and suddenly one die ? Do you have any idea and any advice on how to identify the reset of this Socket disconnection ?

I though about using wireshark but the fact that the application is running on a cluster make it difficult to analyze,sniff the packets remotely. I dont know if i can do it remotely.

Do you have any idea how I could debug this problem?

Thanks

Perhaps an exception got thrown in read(), on the server side, that you didn't catch?