[
https://issues.apache.org/jira/browse/KAFKA-8999?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stefan Hoffmeister updated KAFKA-8999:
--------------------------------------
Description:
The implementation of apache.kafka.common.utils.Utils.delete masks any errors
it encounters during recursive deletes of a directory.
[https://github.com/apache/kafka/blob/2.4/clients/src/main/java/org/apache/kafka/common/utils/Utils.java#L748]
is implemented such that for postVisitDirectory on SimpleFileVisitor any
_unknown_ exception information present is simply discarded.
At the very least, these exceptions should be logged at TRACE level, to be able
to determine the root cause of any such failures.
Beyond logging, it might be worthwhile reviewing whether ignoring an exception
flagged on postVisitDirectory is correct, as that very exception might prevent
the visited directory from being deleted.
This relates indirectly to debugging KAFKA-6647, where the masking of errors
here effectively hides why Kafka Streams sometimes throws
DirectoryNotEmptyException (there is analysis on KAFKA-6647 why this may be the
case - fixing this defect here would at least provide sufficient error context
information)
was:
The implementation of apache.kafka.common.utils.Utils.delete masks any errors
it encounters during recursive deletes of a directory.
[https://github.com/apache/kafka/blob/2.4/clients/src/main/java/org/apache/kafka/common/utils/Utils.java#L748]
is implemented such that for visitFileFailed and postVisitDirectory on
SimpleFileVisitor any _unknown_ exception information present is simply
discarded.
At the very least, these exceptions should be logged at TRACE level, to be able
to determine the root cause of any such failures.
Beyond logging, it might be worthwhile reviewing whether ignoring an exception
flagged on postVisitDirectory is correct, as that very exception might prevent
the visited directory from being deleted.
This relates indirectly to debugging KAFKA-6647, where the masking of errors
here effectively hides why Kafka Streams sometimes throws
DirectoryNotEmptyException (there is analysis on KAFKA-6647 why this may be the
case - fixing this defect here would at least provide sufficient error context
information)
> org.apache.kafka.common.utils.Utils.delete masks errors encountered during
> processing
> -------------------------------------------------------------------------------------
>
> Key: KAFKA-8999
> URL: https://issues.apache.org/jira/browse/KAFKA-8999
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Reporter: Stefan Hoffmeister
> Priority: Major
>
> The implementation of apache.kafka.common.utils.Utils.delete masks any errors
> it encounters during recursive deletes of a directory.
> [https://github.com/apache/kafka/blob/2.4/clients/src/main/java/org/apache/kafka/common/utils/Utils.java#L748]
> is implemented such that for postVisitDirectory on SimpleFileVisitor any
> _unknown_ exception information present is simply discarded.
> At the very least, these exceptions should be logged at TRACE level, to be
> able to determine the root cause of any such failures.
> Beyond logging, it might be worthwhile reviewing whether ignoring an
> exception flagged on postVisitDirectory is correct, as that very exception
> might prevent the visited directory from being deleted.
> This relates indirectly to debugging KAFKA-6647, where the masking of errors
> here effectively hides why Kafka Streams sometimes throws
> DirectoryNotEmptyException (there is analysis on KAFKA-6647 why this may be
> the case - fixing this defect here would at least provide sufficient error
> context information)
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
