[
https://issues.apache.org/jira/browse/SPARK-11078?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andrew Or updated SPARK-11078:
------------------------------
Description:
The new unified memory management model in SPARK-10983 uncovered many brittle
tests that rely on arbitrary thresholds to detect spilling. Some tests don't
even assert that spilling did occur.
We should go through all the places where we test spilling behavior and correct
the tests, a subset of which are definitely incorrect. Potential suspects:
- UnsafeShuffleSuite
- ExternalAppendOnlyMapSuite
- ExternalSorterSuite
- SQLQuerySuite
- DistributedSuite
was:
The new unified memory management model in SPARK-10983 uncovered many brittle
tests that rely on arbitrary thresholds to detect spilling. Some tests don't
even assert that spilling did occur.
We should go through all the places where we test spilling behavior and correct
the tests, a subset of which are definitely incorrect. Potential suspects:
- UnsafeShuffleSuite
- ExternalAppendOnlyMapSuite
- ExternalSorterSuite
- SQLQuerySuite
> Ensure spilling tests are actually spilling
> -------------------------------------------
>
> Key: SPARK-11078
> URL: https://issues.apache.org/jira/browse/SPARK-11078
> Project: Spark
> Issue Type: Sub-task
> Components: Spark Core, Tests
> Reporter: Andrew Or
>
> The new unified memory management model in SPARK-10983 uncovered many brittle
> tests that rely on arbitrary thresholds to detect spilling. Some tests don't
> even assert that spilling did occur.
> We should go through all the places where we test spilling behavior and
> correct the tests, a subset of which are definitely incorrect. Potential
> suspects:
> - UnsafeShuffleSuite
> - ExternalAppendOnlyMapSuite
> - ExternalSorterSuite
> - SQLQuerySuite
> - DistributedSuite
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
