Hello! It is hard to say. Have you tried changing messageQueueLimit? 1024 might be enough for 2 nodes, but with 3 nodes there will be more inter-node communication and you may become stuck on message queues.
It might also help if you collect stack traces once again when under load, so that we could take a look. Regards, -- Ilya Kasnacheev ср, 12 дек. 2018 г. в 09:47, summasumma <bytestre...@gmail.com>: > Hi Ilya, > > The setup is a 3 VM nodes in openstack all connected through each other in > tenant network. > > Following is the configuation used in all 3 ignite nodes to start the > setup: > ignite-ycsb-all-same-network.xml > < > http://apache-ignite-users.70518.x6.nabble.com/file/t2137/ignite-ycsb-all-same-network.xml> > > > Please let me know if the issue can be narrowed down. > Only if i stop one of the ignite nodes out of 3, i get a good OPS > performance. > > Thanks, > ...summa > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >