Thanks very much for your reply. I examined the logfiles again to answer your 
questions:

"[EMAIL PROTECTED]" wrote : 1) You refer to "the master node".  Please confirm 
that this is 62.50.43.211.
  | 

No, at that time the master node was 62.50.43.210. The first logoutput and the 
second one are from this machine, means that the master node (62.50.43.210) 
produced the output "Dead members:0, New members: 0" and immediately after that 
undeployed all the HA-Queues and HA-Topics. Sorry, I should have made that 
clear in my first post.

"[EMAIL PROTECTED]" wrote : 
  | 2) On the node that produced the first bit of logging in your post, do you 
see log entries with this content "New cluster view for partition 
StagePartition: 202" and "New cluster view for partition StagePartition: 201"?
  | 

No, these messages are not present in the logfile.

"[EMAIL PROTECTED]" wrote : 
  | 3) If you have a log entry somewhere that contains "New cluster view for 
partition StagePartition: 200", please compare the list of nodes to the first 
line in the first log entry in your post.  Does it have the same 6 nodes but in 
different order?
  | 

You are right, I can see the same nodes, but in different order

"[EMAIL PROTECTED]" wrote : 
  | What I'm driving at here is I wonder if the machine doing the first bit of 
logging lost a couple view changes, going from 200 to 203.  The result would be 
Dead members:0, New members: 0 but a different order of members.
  | 

Thanks, now I start to understand what is happening. You are right that the 
machine indeed lost some of the view changes, that's a problem I probably have 
to investigate on the network level. 

But the most intersting question for me is: Even if the (Master-)node lost some 
viewchanges,  why does it suddenly undeploy the (HA-)queues and  (HA-)topics? 
And why is the failover not happening, no other node is starting to deploy the 
queues and topics instead. I cannot explain how this is possible and also found 
no information in the docs or in the forums on this issue.

The critical thing is that if I run into this scenario my HA-Queues and 
HA-Topics are not present on any instance, leading to lost messages and 
therefore also lost data. This situation should not be possible at all in a 
cluster. I am not quite sure if this is a cluster issue (I guess so), so if it 
is something related to JMS please let me know so I can ask in JMS-Forum. 

BTW: This is the only real problem we have with the JBoss platform. Everything 
else is working fine and stable. Developing with JBoss really was a breeze, so 
thanks for this great piece of software. 

Thanks again for your help.

Jochen


View the original post : 
http://www.jboss.com/index.html?module=bb&op=viewtopic&p=3954296#3954296

Reply to the post : 
http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=3954296

Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
JBoss-user mailing list
JBoss-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/jboss-user

Reply via email to