Merge Failed error on JBoss 4.0.4RC1 (version 200602071519)

[luism]

I've just installed the new JBoss 4.0.4RC1 and it seems that JBAS-2439 http://jira.jboss.com/jira/browse/JBAS-2439 isn't solved yet.

I started jboss instances on two separate machines waited for the initial joins then unplugged the network cable, I waited again for the member status update and then plugged the network cable and got the following message:

...
22:23:51,518 INFO [DefaultPartition] New cluster view for partition DefaultPartition: 3 ([10.0.1.5:1099, 10.0.1.10:1099] delta: 1)
22:23:51,518 INFO [DefaultPartition] Merging partitions...
22:23:51,519 INFO [DefaultPartition] Dead members: 0
22:23:51,519 INFO [DefaultPartition] Originating groups: [[10.0.1.5:33063 (additional data: 13 bytes)|2] [10.0.1.5:33063 (additional data: 13 bytes)], [Marsk6:33420 (additional data: 14 bytes)|2] [Marsk6:33420 (additional data: 14 bytes)]]
22:23:51,630 ERROR [DefaultPartition] merge failed
java.lang.ClassCastException: java.lang.ClassCastException
at org.jboss.ha.framework.server.DistributedReplicantManagerImpl.mergeMembers(DistributedReplicantManagerImpl.java:890)
at org.jboss.ha.framework.server.DistributedReplicantManagerImpl$MergeMembers.run(DistributedReplicantManagerImpl.java:1033)
...

I've tested with the bundled Jgroups version and with Jgroups 2.2.9.1, the results were the same.
I was hoping that this was solved in this version as it was the reason why I updated from 4.0.3SP1.

If this problem gets confirmed, can it be expected a fix version soon?

Thanks

[brian.stansberry]

What you're reporting is a different issue from JBAS-2439, although very similar. This is the first report I've seen of it; perhaps the similarity in description to JBAS-2439 led people to not report it.

If I can verify it it will be fixed in 4.0.4 FINAL.

[brian.stansberry]

Did you see an error in the logs of your other server, perhaps related to invoking DistributedReplicantManagerImpl.lookupLocalReplicants()? The only way I see to get a ClassCastException at the line indicated is if the remote node threw an exception.

[luism]

I found that the problem was with concurrent.jar, I had replaced the original concurrent.jar with a different version which caused the exception.

With the bundled concurrent.jar all is well.

Thanks