Hi vdpyatkov, Finally, I figured out not only one but all nodes become segmentation due to unknown VMWare infrastructure. I changed failuredetectiontimeout/sockettimeout/networktimeout, and the cluster survived last day, need more time to observe their behavior.
I'm thinking SegmentationPolicy. Why SegmentationPolicy.RESTART_JVM was provided(must work with CommandLineStartup)? Any limitation/implication that soft-restart(in same JVM) won't work? Thanks, Duke -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/How-many-nodes-within-your-ignite-cluster-tp8808p8892.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.