Topic Connections dying with more than 60 nodes

[Aidan <ai...@thalesgroup.com.au>]

Hi guys,

I am currently working on building a computing grid to try to perform many
simulations at once for my PhD thesis. I am using GridGain
(http://www.gridgain.org) for the cloud framework and ActiveMQ 4.1.1 for
messaging (discovery and communication). 

I have managed to get a grid of about 40 nodes to run successfully for a few
hours with no real issues, however when I attempt to add more nodes to the
grid (to about 60 nodes) I lose all TopicConnections and none of my nodes
can reconnect to the broker. 

I am using the default activemq.xml file with the SSL, Stomp and multicast
turned off and am connecting to one known server. When about the 61st node
is added to the grid all TCP communication stops and cannot be reconnected,
however there is no error message printed out on the console. If I restart
the nodes (but not the broker) the console will print out IOExceptions
regarding a Broken Pipe and the clients will not receive any data.
Unfortunately I didn't get a chance to grab any dumps/log files as there are
so many nodes the log files get overwritten pretty quickly.

Is there a configuration option / some design structure I need to be
considering when trying to talk to a cluster of (what will eventually be)
210 nodes? I can't move from ActiveMq 4.1.1 as there appears to be issues
when using the 5.0.* releases with GridGain. I am using dynamicTopic's and
dynamicQueue's and they are set to be not persisted.

I'm pretty sure I've missed something somewhere as it all works fine with 40
nodes, but dies when I get around 60.

Thanks very much,

Aidan




-- 
View this message in context: http://www.nabble.com/Topic-Connections-dying-with-more-than-60-nodes-tp18567995p18567995.html
Sent from the ActiveMQ - User mailing list archive at Nabble.com.