You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Morio Ramdenbourg (JIRA)" <ji...@apache.org> on 2019/03/19 23:16:00 UTC

[jira] [Comment Edited] (HIVE-21480) Fix test TestHiveMetaStore.testJDOPersistanceManagerCleanup

    [ https://issues.apache.org/jira/browse/HIVE-21480?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16796607#comment-16796607 ] 

Morio Ramdenbourg edited comment on HIVE-21480 at 3/19/19 11:15 PM:
--------------------------------------------------------------------

Update: the current implementation of the test does nothing. The getJDOPersistenceManagerCacheSize method was always returning -1, since the fields in ObjectStore had been updated, and it could not grab the actual size of the PersistenceManagerCache.


was (Author: mramdenbourg):
Update: the test was testing nothing. The getJDOPersistenceManagerCacheSize method was always returning -1, since the fields in ObjectStore had been updated, and it could not grab the actual size of the PersistenceManagerCache.

> Fix test TestHiveMetaStore.testJDOPersistanceManagerCleanup
> -----------------------------------------------------------
>
>                 Key: HIVE-21480
>                 URL: https://issues.apache.org/jira/browse/HIVE-21480
>             Project: Hive
>          Issue Type: Improvement
>          Components: Standalone Metastore
>    Affects Versions: 4.0.0
>            Reporter: Morio Ramdenbourg
>            Assignee: Morio Ramdenbourg
>            Priority: Major
>
> [TestHiveMetaStore#testJDOPersistanceManagerCleanup|https://github.com/apache/hive/blob/master/standalone-metastore/metastore-server/src/test/java/org/apache/hadoop/hive/metastore/TestHiveMetaStore.java#L3140-L3162] tests whether the JDO persistence manager cache cleanup was performed correctly when a HiveMetaStoreClient executes an API call, and closes. It does this by ensuring that the cache object count before the API call, and after closing, are the same. However, there are some assumptions that are not always correct, and can cause flakiness.
> For example, lingering resources could be present from previous tests or from setup depending on how PTest runs it, and can cause the object count to sometimes be incorrect. We should rewrite this test to account for this flakiness that can occur.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)