[
https://issues.apache.org/jira/browse/MAPREDUCE-7524?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Paco Chan updated MAPREDUCE-7524:
---------------------------------
Description:
{code:java}
org.apache.hadoop.mapreduce.lib.input.TestCombineFileInputFormat.testSplitPlacementForCompressedFiles{code}
makes a number of assertions that are not guaranteed by internal
implementation.
For example, the test makes assumptions on the order in which racks are
processed. Splits in the HDFS can appear in a different order depending on the
internal state / implementation / metadata which is not guaranteed by the
interface.
For example, there are assertions such as split.equals(splits.get(0)), which
assumes a specific ordering of InputSplit objects that may vary depending on
internal implementation.
This can be addressed by eliminating assertions based on the internal
implementation / ways that the racks are split.
was:
org.apache.hadoop.mapreduce.lib.input.TestCombineFileInputFormat.testSplitPlacementForCompressedFiles
makes a number of assertions that are not guaranteed by internal
implementation.
For example, the test makes assumptions on the order in which racks are
processed. Splits in the HDFS can appear in a different order depending on the
internal state / implementation / metadata which is not guaranteed by the
interface.
For example, there are assertions such as split.equals(splits.get(0)), which
assumes a specific ordering of InputSplit objects that may vary depending on
internal implementation.
This can be addressed by eliminating assertions based on the internal
implementation / ways that the racks are split.
> testSplitPlacementForCompressedFiles is non-deterministic from time to time
> ---------------------------------------------------------------------------
>
> Key: MAPREDUCE-7524
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-7524
> Project: Hadoop Map/Reduce
> Issue Type: Test
> Reporter: Paco Chan
> Priority: Trivial
>
> {code:java}
> org.apache.hadoop.mapreduce.lib.input.TestCombineFileInputFormat.testSplitPlacementForCompressedFiles{code}
> makes a number of assertions that are not guaranteed by internal
> implementation.
>
> For example, the test makes assumptions on the order in which racks are
> processed. Splits in the HDFS can appear in a different order depending on
> the internal state / implementation / metadata which is not guaranteed by the
> interface.
>
> For example, there are assertions such as split.equals(splits.get(0)), which
> assumes a specific ordering of InputSplit objects that may vary depending on
> internal implementation.
>
> This can be addressed by eliminating assertions based on the internal
> implementation / ways that the racks are split.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
