[
http://jira.jboss.com/jira/browse/JGRP-655?page=comments#action_12397441 ]
Vladimir Blagojevic commented on JGRP-655:
------------------------------------------
Resolved on HEAD:
Revision Changes Path
1.168 +46 -64 JGroups/src/org/jgroups/JChannel.java
Resolved on 2.6:
1.158.2.8 +46 -64 JGroups/src/org/jgroups/JChannel.java
Channel#connect should stop stack in case it failed
---------------------------------------------------
Key: JGRP-655
URL:
http://jira.jboss.com/jira/browse/JGRP-655
Project: JGroups
Issue Type: Bug
Affects Versions: 2.4, 2.5, 2.6
Reporter: Bela Ban
Assigned To: Bela Ban
Fix For: 2.7, 2.6.2
[Vladimir]
I think I got one of the last failures in modified ViewDeliveryDemo. I am testing 2.6
branch on atlanta cluster. It was related to a case where no traffic seems to be going
between members after connect failure occurred. A connect failure specifically related to
join reject due to a member already being included in a view - this was not port related I
believe but rather a race of sending join request. I used port manager.
Anyhow, to be honest I did not understand the underlying reasons but I did something that
seems to be intuitive. In connect method I reversed stack to original state in case of
connect failure. That includes calling stopStack and init().
if(res != null && res instanceof Exception) {
try {
prot_stack.stopStack();
}
catch(Exception e) {
if(log.isErrorEnabled()) log.error("exception: " + e);
} init();
throw new ChannelException("connect() failed", (Throwable)res);
}
}
After this fix I did not see the problem again an I ran three member test for about two
hours.
Cheers.
--
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