[jboss-jira] [JBoss JIRA] Created: (JGRP-259) Unidentified errors with TCP/TCPPING/MERGE2 protocol stack.

[Tomasz Skutnik (JIRA)]

Unidentified errors with TCP/TCPPING/MERGE2 protocol stack.
-----------------------------------------------------------

                 Key: JGRP-259
                 URL: http://jira.jboss.com/jira/browse/JGRP-259
             Project: JGroups
          Issue Type: Bug
    Affects Versions: 2.2.9.2
         Environment: Debian GNU/Linux 2.4.27, JDK 1.5.0_05
            Reporter: Tomasz Skutnik
         Assigned To: Bela Ban
         Attachments: Test1.java

I've encountered connect/disconnect problems during my experiments with
TCP/TCPPING/MERGE2 protocol stack. I run client program (see attachment)
on 2 different hosts (witnin two VMWare virtual hosts). One of the hosts
(h1) plays role of well-known host from which initial membership will be
retrieved, the other plays role of plain client.

I use following command lines to run client programs:

- on host h1:
$ java -cp \
  jgroups-all-2.2.9.2.jar:concurrent.jar:commons-logging-1.1.jar:\
  classes test.Test1 "h1" "h1[7800]"

- on host h2
$ java -cp \
  jgroups-all-2.2.9.2.jar:concurrent.jar:commons-logging-1.1.jar:\
  classes test.Test1 "h2" "h1[7800]"

When I run client programs in h1, h2 order everyting works fine. But if
I run them in h2, h1 order, or if I kill h1 and then rerun it something
weird happens: merge of two disjoint groups happens (MergeView is
transmitted), but then followin exception occurs:

2006-07-10 12:01:32 org.jgroups.Message readHeader
SEVERE: magic number 306192 is not available in magic map
2006-07-10 12:01:32 org.jgroups.protocols.TP handleIncomingPacket
SEVERE: failed unmarshalling message
java.io.IOException: failed read header: java.lang.NullPointerException
        at org.jgroups.Message.readHeader(Message.java:697)
        at org.jgroups.Message.readFrom(Message.java:614)
        at org.jgroups.protocols.TP.bufferToMessage(TP.java:974)
        at org.jgroups.protocols.TP.handleIncomingPacket(TP.java:830)
        at org.jgroups.protocols.TP.receive(TP.java:781)
        at org.jgroups.protocols.TCP.receive(TCP.java:226)
        at
org.jgroups.blocks.ConnectionTable.receive(ConnectionTable.java:471)
        at
org.jgroups.blocks.ConnectionTable$Connection.run(ConnectionTable.java:813)
        at java.lang.Thread.run(Thread.java:595)

Because there's no magic number in magic map (whatever that means) it
causes NPE within org.jgroups.Message class (line 687).

>From now on everything is going downhill - jgroups logs infinite stream
of retransmition errors, e.g.:

SEVERE: magic number 306192 is not available in magic map
2006-07-10 12:01:54 org.jgroups.protocols.pbcast.NAKACK handleXmitRsp
SEVERE: message did not contain a list (LinkedList) of retransmitted
messages: java.io.IOException: failed read header:
java.lang.NullPointerException

This flow of exceptions does not stop until all clients (h1 and h2) are
killed - stopping h1 does not stop exceptions on h2 and vice versa.



-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira