[
https://issues.apache.org/jira/browse/HIVE-18670?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16387373#comment-16387373
]
Scott Eade commented on HIVE-18670:
-----------------------------------
{code:java}
-- Example 3
> CREATE DATABASE my_important_database LOCATION /hive/my_important_database;
-- Add a bunch of tables and data.
-- Now create a second database in the same location.
> CREATE DATABASE trouble_maker LOCATION /hive/my_important_database;
-- This will succeed and delete the important data.
> DROP DATABASE trouble_maker;{code}
This is based on a real example. HDFS snapshots and distcp backups to the
rescue.
> Prevent DROP DATABASE If Other Data Exists
> ------------------------------------------
>
> Key: HIVE-18670
> URL: https://issues.apache.org/jira/browse/HIVE-18670
> Project: Hive
> Issue Type: Improvement
> Components: HiveServer2
> Affects Versions: 3.0.0, 2.3.2
> Reporter: BELUGA BEHR
> Priority: Major
>
> A user is not able to drop a database if it has tables under it unless they
> include the _CASCADE_ keyword to their DROP DATABASE statement.
> [https://cwiki.apache.org/confluence/display/Hive/LanguageManual+DDL#LanguageManualDDL-DropTable]
> I would like to propose that, if {{hive.mapred.mode}} is set to 'strict',
> Hive also checks for other data before dropping the database.
> For example, if the database is stored within HDFS, then Hive should check if
> there exists other data, not even necessarily related to Hive, within the
> database's HDFS directory, before dropping.
> The examples are:
> {code:java|title=Example 1}
> /hive/my_database
> /hive/my_database/my_table
> -- Does not succeed because 'my_table' exists
> > DROP DATABASE my_database;
> -- Succeeds and removes the root directory /hive/my_database
> > DROP DATABASE my_database CASCADE;
> {code}
> {code:java|title=Example 2}
> /hive/my_database
> /hive/my_database/my_important_file.txt
> -- Succeeds because no tables exist, but I just lost my "important" file
> > DROP DATABASE my_database;
> {code}
> This "feature" is just to prevent people from shooting themselves in the
> foot, even if they shouldn't be using Hive space for storing unrelated data.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
