Hi Aleksey, In order to ensure that a transaction is rolled back during a node failure, you need to orchestrate messages in such a way that at least one participating node does not receive a prepare request. As an example, you can look at org.apache.ignite.internal.processors.cache.distributed.dht.IgniteCachePrimaryNodeFailureRecoveryAbstractTest.
--AG 2017-03-09 20:09 GMT+03:00 ALEKSEY KUZNETSOV <alkuznetsov...@gmail.com>: > Hi all! Is it possible to test on multiple jvm's ? Let's say, to have > multiple ignite instances deployed on different jvms ? If yes, how can i > create them and address them in code ? > -- > > *Best Regards,* > > *Kuznetsov Aleksey* >