hivemetastore memory leak with ranger

[#1]

hive metastore is consuming complete heap(tried max heap : 24G) after sometime  the metastore is not responding , the metastore is ok if we disable ranger hive authentication  this issue is reported in (https://jira.apache.org/jira/browse/HIVE-20568) ,

It seems ranger plugin is causing memory leaks during audit when it encounters hive db names containing '_'

just wondering to know if there is a work around for it in HDP-3.1.0 or the upgrade from HDP-3.1 to HDP-3.1.4 will fix it,  Thanks in advance.

Current Version  and config Information :

========================

HDP-3.1

hive-3.1

Hive Authentication - Ranger

JDK-1.8

mysql -7.5

mysql-connector-8.0.13.jar

[#2]

@nagaiik  We dont have ranger plugin for Hive metastore. Ranger plugin is configured for Hive Server2 and runs as part of HS2 JVM. 

If it is Hive metastore that hangs then it is not related to ranger. Heap dump of the process must be analyzed  to identify what is causing high memory usage. 

HIVE-20568 is the fix to not convert special characters in the table/db names. Due to the conversion of names, ranger authorization was not enforced as names are not same as the one set in policies. This doesnt have any workaround can be fixed on with upgrade to HDP-3.1.4 or contact cloudera hotfix. Refer the release notes for fixed issues. 

https://docs.cloudera.com/HDPDocuments/HDP3/HDP-3.1.4/release-notes/content/fixed_issues.html

[#3 · Accepted solution]

Hi @rguruvannagari 

Thanks for alot  for the reply , not sure if heap space is filled  during compaction or ranger hive audit if we set  hive authentication to none then it is ok , please see the following issue.

https://community.cloudera.com/t5/Support-Questions/hive-metastore-is-not-responding-but-alive-with-...

Thanks 

Nag