Hi, Joe, >I double checked the config I posted, the one (I think) I'm currently running and the one I send with the logs and they all > seem to be the same Sorry , this is my mistake -- I picked up wrong config for analysis. Please disregard my previous suggestions.
>>Each ignite node (i.e. each ignite.sh run) is on its own dedicated >> physical hardware (2x Intel Xeon w/ 8 cores each..so /proc/cpuinfo shows >> 32 >> processors, all have 64GB RAM some have 128GB, *no local disk* In the paragraph above you said your nodes have "no local disk" -- does that mean that all them use a common shared file system? I wonder, if that may cause a problem due to some slow file system access. Overall, it seems that the issue is not anyhow related to IGFS / map-reduce, and caused by an inter-node communication problem. Can you please create a ticket in IGNITE Jira on that and attach the most detailed stack trace lying under "Closing NIO session because unhandled exception" error. -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Help-with-tuning-for-larger-clusters-tp1692p1813.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.