OutOfMemory exception on a 2 node cluster freezes the cluste

[lordbritish]

I have a 2 node cluster - one of the nodes has more system memory than the other one.

Today, the node with the less memory threw a OutOfMemory exception. The result was that the entire cluster became unresponsive even though the node with more memory had enough available memory.

The desired behavior is of course for the node that is not out of memory to recieve the requests.

This was not happening.

I brought down the node that was out of memory and everything was ok after that - requests were being honored by the remaining node.

This is with JBoss 3.2.3

Please investigate and fix.

[slaboure]

> "Please investigate and fix."

You probably meant "Please investigate and fix IMMEDIATELY.", right ?

[kosulin]

Mr. LordBritish:
Would you please be so kind to create a bug report in Bugzilla:
https://sourceforge.net/tracker/?group_id=22866&atid=376685

And, please, include some additional information: amount of memory, hardware platform and operating system details, and jvm runtime parameters for both nodes, JBoss version, cluster configuration parameters, etc.? JBoss log messages could be helpful also. Then Sacha can investigate and fix it ;-)

Vlad

P.S. Sacha, where is emoticon?
;-)

[belaban]

This is probably due to you using the default configuration and banging the system with messages. In this case, I suggest you use the high-performance (flow-control) based configuration that I will add to JBoss within the next couple of weeks. In the meantime you can check out JGroups/conf/fc.xml or fc-fast.xml, and convert them to the format required by cluster-service.xml yourself.


Bela