You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Sankar Hariappan (JIRA)" <ji...@apache.org> on 2017/02/16 10:54:41 UTC

[jira] [Comment Edited] (HIVE-15910) Improvements in Hive Unit Test by using In-memory Derby DB

    [ https://issues.apache.org/jira/browse/HIVE-15910?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15869432#comment-15869432 ] 

Sankar Hariappan edited comment on HIVE-15910 at 2/16/17 10:54 AM:
-------------------------------------------------------------------

[~wzheng] I was able to reproduce some of TestWorker failures locally and noticed that the failures are happening when you run the same test case twice where the second execution sees non-empty table's directory under compactor_test_tables. There is DELETEONEXIT hook added which doesn't work for some reason. The double run of test cases could be due to TestWorker2 which extends TestWorker.
Fixed this issue by adding directory cleanup before creating new table. Patch-05 is provided with this change.
Please have a check.


was (Author: sankarh):
Fixed some cleanup issues in TestWorker test cases.

> Improvements in Hive Unit Test by using In-memory Derby DB
> ----------------------------------------------------------
>
>                 Key: HIVE-15910
>                 URL: https://issues.apache.org/jira/browse/HIVE-15910
>             Project: Hive
>          Issue Type: Test
>          Components: Tests
>            Reporter: Sankar Hariappan
>            Assignee: Wei Zheng
>         Attachments: HIVE-15910.01.patch, HIVE-15910.05.patch, HIVE-15910.2.patch, HIVE-15910.3.patch, HIVE-15910.4.patch
>
>
> Hive UT currently uses Derby DB with storage on disk which have some practical problems.
> 1. The run-time of Hive unit tests are high as need to operate on the disk quite often.
> 2. It can cause conflict if multiple test cases operates on the same table name (such as table being created already exist).
> To solve these problems, we shall use an in-memory storage option of Derby DB which can be even persisted if the test case demands that.
> https://db.apache.org/derby/docs/10.8/devguide/cdevdvlpinmemdb.html



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)