[
https://issues.apache.org/jira/browse/HIVE-28588?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17984954#comment-17984954
]
Zsolt Miskolczi commented on HIVE-28588:
----------------------------------------
I cannot recall it, it is out of my brain memory limit. Based on the problem
and the change, I assume because that PR renames the JDBC table:
[https://github.com/apache/hive/pull/4913/files#diff-0d75cb0cb4d9e8d36c7ecb7ac39ad7216786d4016d80c6168b8c6544fcafb895]
That also means, the JDBC table is not reused between test cases so it should
fix the problem.
As far I remember, the root cause for that kind of problem is because there is
no automatic cleanup in the Derby database after this execution - so the tests
are passing if different split runs them. And fail otherwise.
If we make the table names unique across tests, it fixes the issue in one way.
> JDBC table related flakiness
> ----------------------------
>
> Key: HIVE-28588
> URL: https://issues.apache.org/jira/browse/HIVE-28588
> Project: Hive
> Issue Type: Bug
> Components: HiveServer2
> Reporter: Zsolt Miskolczi
> Priority: Major
> Fix For: Not Applicable
>
>
> There are couple of test failures on downstream, like:
> authorization_privilege_objects.q, external_jdbc_rowcount.q
> The pattern is similar:
> * Create a Derby db using GenericUDFDBOutput
> * Create an external table on top of that using
> org.apache.derby.jdbc.EmbeddedDriver
> Creating the Derby db returns with an error: table already exist. So that
> GenericUDFDBOutput returns with 1 instead of 0. And the qtest fails with diff
> in output.
> Example:
> HS2 log:
> {noformat}
> 2024-10-21T05:51:51,249 ERROR [1d2e8bd2-f184-472a-8f9a-44daabc9dd36 main]
> example.GenericUDFDBOutput: Underlying SQL exception
> java.sql.SQLException: Table/View 'SIMPLE_DERBY_TABLE1' already exists in
> Schema 'APP'.{noformat}
> Test output:
> {noformat}
> Client Execution succeeded but contained differences (error code = 1) after
> executing authorization_privilege_objects.q
> 246c246
> < 1
> ---
> > 0{noformat}
>
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
