Connection was forcibly closed by the remote host

Question

I have a java.nio.channels.SocketChannel in my jSCSI implamantation that is disconnecting when I try to open a driver with size greater than 4GB. The iscsi RFC says that the BasicHeaderSegment.BHS_FIXED_SIZE may be 48, so with this position I can read the bytes on the channel. The java doc says 5 types of errors, but my app is throwing the last one that doesn't say a specific information.

1. NotYetConnectedException
2. ClosedChannelException
3. AsynchronousCloseException
4. ClosedByInterruptException
5. IOException

Code:

public final int read(final SocketChannel sChannel) throws InternetSCSIException, IOException, DigestException {
// read Basic Header Segment first to determine the total length of this
// Protocol Data Unit.
clear();

final ByteBuffer bhs = ByteBuffer.allocate(BasicHeaderSegment.BHS_FIXED_SIZE);
int len = 0;
while (len < BasicHeaderSegment.BHS_FIXED_SIZE) {
    int lens = sChannel.read(bhs);
    if (lens == -1) {
        // The Channel was closed at the Target (e.g. the Target does
        // not support Multiple Connections)
        // throw new ClosedChannelException();
        return lens;
    }
    len += lens;
    LOGGER.trace("Receiving through SocketChannel: " + len + " of maximal " + BasicHeaderSegment.BHS_FIXED_SIZE);

}
bhs.flip();

error:

java.io.IOException: An existing connection was forcibly closed by the remote host
    at sun.nio.ch.SocketDispatcher.read0(Native Method)
    at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:43)
    at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
    at sun.nio.ch.IOUtil.read(IOUtil.java:197)
    at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:379)
    at org.jscsi.parser.ProtocolDataUnit.read(ProtocolDataUnit.java:417)
    at org.jscsi.target.connection.TargetSenderWorker.receiveFromWire(TargetSenderWorker.java:145)
    at org.jscsi.target.connection.Connection$TargetConnection.receivePdu(Connection.java:217)
    at org.jscsi.target.connection.phase.TargetFullFeaturePhase.execute(TargetFullFeaturePhase.java:96)
    at org.jscsi.target.connection.Connection$TargetConnection.call(Connection.java:264)
    at org.jscsi.target.connection.Connection$TargetConnection.call(Connection.java:79)
    at java.util.concurrent.FutureTask.run(FutureTask.java:262)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
    at java.lang.Thread.run(Thread.java:744)
[pool-9-thread-1] INFO org.jscsi.target.connection

Thanks in advance, Felipe

Solution

Your problem description is incorrect. The SocketChannel isn't closed: the connection is; and it isn't happening when you try to open the connection: it is happening when you read from it.

This usually results from sending something after the peer has already closed the connection, which in turn usually means you had previously sent it something it didn't understand.

Comments

[user207421]

Note the correction to your title. Your SocketChannel is still open. It is the connection that was forcibly closed.

[Felipe Gutierrez]

I am searching on this link the properties of iscsi, maybe I can increase the timeout, but I dont know what to edit blogs.msdn.com/b/san/archive/2008/07/27/…

[user207421]

Timeouts have nothing to do with it. You don't seem to have read this answer.

[Felipe Gutierrez]

I got to mount drivers multiplied for 2. eg: 4Gb, 8GB, 16GB

[Edward Tomasz Napierala]

You mean "drive", eg. a LUN, disk, right? So, the 4GB looks suspicious. There is nothing in iSCSI itself that could have an edge condition related to 4GB. Perhaps it's a SCSI-level problem? Again, anything interesting in target logs?