[
https://issues.apache.org/jira/browse/HIVE-16950?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16326165#comment-16326165
]
Oleksiy Sayankin commented on HIVE-16950:
-----------------------------------------
Hi all.
Let's discuss some ideas about this issue. When I have implemented verification
for tables (check if table folder exists before creation of a table, and if yes
- throw an exception), I have faced some problems here:
# SemanticAnalyzer uses temp tables placed in one folder, so it starts to fail
in INSERT operations.
# Customers may want to create external tables with existing data/folders to
escape from data copying.
So I have added two exceptions: temporary and external tables may use the same
folder. Even in that case I have some tests failed (e.g.
org.apache.hive.hcatalog.api.repl.commands.TestCommands.testBasicReplEximCommands).
Import command does not work with the fix for tables:
{code:java}
hive> export table student to '/user/mysudent';
hive> import table imported_student from '/user/mysudent';
{code}
results with
{code:java}
FAILED: Execution Error, return code 1 from
org.apache.hadoop.hive.ql.exec.DDLTask. InvalidObjectException(message:Failed
to create table. Table directory already exists
{code}
So my suggestion is to implement fix for databases but not for the tables. Any
ideas?
> Dropping hive database/table which was created explicitly in default database
> location, deletes all databases data from default database location
> -------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HIVE-16950
> URL: https://issues.apache.org/jira/browse/HIVE-16950
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Affects Versions: 1.2.1
> Reporter: Rahul Kalgunde
> Assignee: Oleksiy Sayankin
> Priority: Minor
> Attachments: HIVE-16950.1.patch, HIVE-16950.2.patch
>
>
> When database/table is created explicitly pointing to the default location,
> dropping the database/table deletes all the data associated with the all
> databases/tables.
> Steps to replicate:
> in below e.g. dropping table test_db2 also deletes data of test_db1 where as
> metastore still contains test_db1
> hive> create database test_db1;
> OK
> Time taken: 4.858 seconds
> hive> describe database test_db1;
> OK
> test_db1
> hdfs://sandbox.hortonworks.com:8020/apps/hive/warehouse/test_db1.db root
> USER
> Time taken: 0.599 seconds, Fetched: 1 row(s)
> hive> create database test_db2 location '/apps/hive/warehouse' ;
> OK
> Time taken: 1.457 seconds
> hive> describe database test_db2;
> OK
> test_db2
> hdfs://sandbox.hortonworks.com:8020/apps/hive/warehouse root USER
> Time taken: 0.582 seconds, Fetched: 1 row(s)
> hive> drop database test_db2;
> OK
> Time taken: 1.317 seconds
> hive> dfs -ls /apps/hive/warehouse;
> ls: `/apps/hive/warehouse': No such file or directory
> Command failed with exit code = 1
> Query returned non-zero code: 1, cause: null
> hive> describe database test_db1;
> OK
> test_db1
> hdfs://sandbox.hortonworks.com:8020/apps/hive/warehouse/test_db1.db root
> USER
> Time taken: 0.629 seconds, Fetched: 1 row(s)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
