[
https://issues.apache.org/jira/browse/HIVE-29408?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18055493#comment-18055493
]
Stamatis Zampetakis commented on HIVE-29408:
--------------------------------------------
With HIVE-29401 in place there is no noticeable performance gain between
in-memory and file-system Derby databases. Since persisted databases can be
easier inspected if needed and there is no obvious benefit in getting this
change in I will close this as wont'fix. It's not worth risking destabilizing
the CI which works rather fine at the moment since using in-memory databases
may increase the memory footprint and lead to other issues. If in the future we
find more compelling reasons to use in-memory databases in qtests we can revive
this ticket and the respective PR.
> Use in-memory database in tests using qt:database:derby option
> --------------------------------------------------------------
>
> Key: HIVE-29408
> URL: https://issues.apache.org/jira/browse/HIVE-29408
> Project: Hive
> Issue Type: Task
> Components: Test
> Reporter: Stamatis Zampetakis
> Assignee: Stamatis Zampetakis
> Priority: Major
> Labels: pull-request-available
>
> There are many JDBC storage handler tests that create and use Derby databases
> by using the qt:database:derby option. The databases are currently persisted
> in the file-system but this can significantly slow-down the tests especially
> in environments (CI) where I/O operations are slow.
> Derby supports creating fully in-memory databases that are not bound to the
> I/O performance of the underlying filesystem. The goal of this ticket is to
> use in-memory Derby databases whenever the qt:database:derby option is used
> in tests.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
