[ https://issues.jboss.org/browse/ISPN-4037?page=com.atlassian.jira.plugin.... ] Jakub Markos updated ISPN-4037: ------------------------------- Attachment: serverlogs.zip Here are a few more details about the test: 2 nodes, replicated cache with replication-queue size 3. We do 2 puts and verify that there are 2 entries in the cluster via the JMX numberOfEntries from both nodes (this may fail), and then 3 more puts so that the queue is filled at some node, and verify that the number of entries is 8 (3 entries are replicated, so 3x2, plus 2 unreplicated). The reason it may fail is because the put counts towards the numberOfEntries statistic only if the put came to the node which is the lock owner for the key. If a node receives a new key for which it's not an owner, it simply puts it in the replication queue, and if it is the lock owner, it adds the entry and then puts it in the queue (so it reports it in the numberOfEntries). The test passes if we're lucky and all puts go directly to the lock owners. The attached logs are from a run with 5 puts, where the first node is the lock owner for all keys, but 2 puts go the second node so they are not reported in numberOfEntries. I'll fix the test so it only assumes queueSize x 2 number of entries after the replication. -- This message was sent by Atlassian JIRA (v6.3.1#6329)