[
https://issues.apache.org/jira/browse/SPARK-34077?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Adam Hamlin updated SPARK-34077:
--------------------------------
Description:
Basically, spinning down a SparkSession using session.stop() does not appear to
clean up JDBC connection pool daemon threads. Please see this StackOverflow
question for more info (and a code snippet to reproduce):
[https://stackoverflow.com/questions/65428297/spark-background-threads-not-cleaned-up-when-using-embedded-derby-metastore]
As mentioned in the question, this can eat up resources during unit tests where
sessions are spun up/spun down throughout a test suite.
I haven't tested this against a non-embedded metastore, but this does occur
when using derby in memory, too.
was:
Basically, spinning down a SparkSession using session.stop() does not appear to
clean up JDBC connection pool daemon threads. Please see this StackOverflow
question for more info (and a code snippet to reproduce):
[https://stackoverflow.com/questions/65428297/spark-background-threads-not-cleaned-up-when-using-embedded-derby-metastore]
As mentioned in the question, this can eat up resources during unit tests where
sessions are spun up/spun down throughout a test suite.
I haven't tested this against a non-embedded metastore, but this does occur
when using derby in memory.
> Metastore daemon threads not cleaned up with session.stop()
> -----------------------------------------------------------
>
> Key: SPARK-34077
> URL: https://issues.apache.org/jira/browse/SPARK-34077
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 2.4.3, 3.0.0
> Reporter: Adam Hamlin
> Priority: Minor
>
> Basically, spinning down a SparkSession using session.stop() does not appear
> to clean up JDBC connection pool daemon threads. Please see this
> StackOverflow question for more info (and a code snippet to reproduce):
> [https://stackoverflow.com/questions/65428297/spark-background-threads-not-cleaned-up-when-using-embedded-derby-metastore]
> As mentioned in the question, this can eat up resources during unit tests
> where sessions are spun up/spun down throughout a test suite.
> I haven't tested this against a non-embedded metastore, but this does occur
> when using derby in memory, too.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
