[
https://issues.apache.org/jira/browse/HIVE-22512?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ashutosh Bapat updated HIVE-22512:
----------------------------------
Attachment: HIVE-22512.03.patch
Status: Patch Available (was: In Progress)
Patch with Mahesh's comments addressed.
> Use direct SQL to fetch column privileges in refreshPrivileges
> --------------------------------------------------------------
>
> Key: HIVE-22512
> URL: https://issues.apache.org/jira/browse/HIVE-22512
> Project: Hive
> Issue Type: Improvement
> Components: Metastore
> Affects Versions: 4.0.0
> Reporter: Ashutosh Bapat
> Assignee: Ashutosh Bapat
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-22512.01.patch, HIVE-22512.02.patch,
> HIVE-22512.03.patch
>
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> refreshPrivileges() calls listTableAllColumnGrants() to fetch the column
> level privileges. The later function retrieves the individual column objects
> by firing one query per column privilege object, thus causing the backend db
> to be swamped by these queries when PrivilegeSynchronizer is run.
> PrivilegeSynchronizer synchronizes privileges of all the databases, tables
> and columns and thus the backend db can get swamped really bad when there are
> thousands of tables with hundreds of columns.
> The output of listTableAllColumnGrants() is not used completely so all the
> columns the PM has tried to retrieves anyway goes waste.
> Fix this by using direct SQL to fetch column privileges.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
