[ https://issues.jboss.org/browse/ISPN-2750?page=com.atlassian.jira.plugin.... ] Dan Berindei resolved ISPN-2750. -------------------------------- Resolution: Won't Fix Looks like a configuration problem again: numSegments is only 40, and there are 32 nodes, which means the segments are not evenly divided between the cache members. Here's an ASCII "graph" to that shows how many segments are owned by each node in a sample consistent hash ('=' means it's a primary owner, '+' means it's a backup owner): {noformat} + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + = + + = = = = + = + + + + = + + + + + + + + = + + + + + + + + + = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = {noformat} You can see from the graph that before the ISPN-2643 fix, when the HotRod client would contact a random owner, the load was balanced more evenly. After that fix however, the HotRod client only contacts the primary owner, and there is a clear difference in load between the nodes who primary-own 2 segments and the nodes who primary-own only 1 segment. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira