[ https://issues.apache.org/jira/browse/CASSANDRA-6320?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13827513#comment-13827513 ]
Jeremy Hanna commented on CASSANDRA-6320: ----------------------------------------- It was reproduced in 1.2.6 with a handful of backported patches (the version is 1.2.6.1). > During range movements, nodetool ring reports incorrect replicas > ---------------------------------------------------------------- > > Key: CASSANDRA-6320 > URL: https://issues.apache.org/jira/browse/CASSANDRA-6320 > Project: Cassandra > Issue Type: Bug > Components: Core > Reporter: Jeremy Hanna > Priority: Minor > > There are two datacenters and the keyspace configuration for this keyspace is > set to 3 in each datacenter. During a rebalance of the ring, during the > range movements, nodetool ring <keyspace> is reporting that 2 replicas are in > the datacenter where movement is occurring. > NodeCmd calls StorageService.effectiveOwnership which seems like it should be > definitive in terms of calculating the number of replicas. Then > NodeCmd.printRing goes through that and takes each filtered list of > datacenters and prints each with replica and node information via > NodeCmd.printDc. > Hopefully it's just a corner case bug in the reporting itself, but I thought > I would log this in case there was a corner case with the actual ownership > during range movements. Also the range movements could be a red herring that > is just coincidental to this problem. -- This message was sent by Atlassian JIRA (v6.1#6144)