[ https://issues.apache.org/jira/browse/HIVE-12250?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14975704#comment-14975704 ]
Naveen Gangam commented on HIVE-12250: -------------------------------------- The failed tests do not appear to be related to the change. > Zookeeper connection leaks in Hive's HBaseHandler. > -------------------------------------------------- > > Key: HIVE-12250 > URL: https://issues.apache.org/jira/browse/HIVE-12250 > Project: Hive > Issue Type: Bug > Components: HiveServer2 > Affects Versions: 1.1.0 > Reporter: Naveen Gangam > Assignee: Naveen Gangam > Attachments: HIVE-12250.2.patch, HIVE-12250.patch > > > HiveServer2 performance regresses severely due to what appears to be a leak > in the ZooKeeper connections. lsof output on the HS2 process shows about 8000 > TCP connections to the ZK ensemble nodes. > grep TCP lsof-hive-node11 | grep node11 | grep -E "node03|node04|node05" | wc > -l > 7866 > grep TCP lsof-hive-node11 | grep node11 | grep -E "node03" | wc -l > 2615 > grep TCP lsof-hive-node11 | grep node11 | grep -E "node04" | wc -l > 2622 > grep TCP lsof-hive-node11 | grep node11 | grep -E "node05" | wc -l > 2629 > node11 - HMS node > node03, node04 and node05 are the hosts for zookeeper ensemble. -- This message was sent by Atlassian JIRA (v6.3.4#6332)