[ https://jira.jboss.org/jira/browse/JGRP-1027?page=com.atlassian.jira.plug... ] Bela Ban updated JGRP-1027: --------------------------- Fix Version/s: 2.9 -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JGRP-1027?page=com.atlassian.jira.plug... ] Gray Watson updated JGRP-1027: ------------------------------ Attachment: NAKACKEarlyTest.java Here's a unit test which demonstrates the NAKACK missing window problem. I wrote the test to demonstrate the issue but I've seen the issue in production on our systems which use multicast messaging a lot. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JGRP-1027?page=com.atlassian.jira.plug... ] Gray Watson commented on JGRP-1027: ----------------------------------- That patch may be in the wrong place. It should be in the send() method or at least coming down the stack -- not up. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JGRP-1027?page=com.atlassian.jira.plug... ] Bela Ban commented on JGRP-1027: -------------------------------- The unit test is flawed, it sends a message in viewAccepted() using Channel.down() instead of Channel.send(). Had you used Channel.send(), you'd get an unconnected exception. When viewAccepted is called, the current channel may or may not yet be connected (if this is the first view), so you cannot depend on being able to send a message, the javadocs for viewAccepted() is: /** * Called when a change in membership has occurred. * <b>No long running actions or sending of messages should be done in this callback.</b> * If some long running action needs to be performed, it should be done in a separate thread.<p/> * Note that on reception of the first view (a new member just joined), the channel will not yet be * in the connected state. This only happens when {@link Channel#connect(String)} returns. */ void viewAccepted(View new_view); Setting an initial seqno to the default (1) is dangerous, as the seqnos are only set via initial digests. So I don't know what the ramifications are, but I certainly don't want to change that code and possibly introduce subtle bugs. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://jira.jboss.org/jira/browse/JGRP-1027?page=com.atlassian.jira.plug... ] Gray Watson commented on JGRP-1027: ----------------------------------- That's fine but we have seen this in production. The test case simulated the error condition and you can say that it is flawed but I didn't write it for amusement purposes but to simulate what we had seen previously. If it reoccurs or if I have any additional information, I'll reopen the issue. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira