[
https://issues.apache.org/jira/browse/NIFI-6846?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16968419#comment-16968419
]
ASF subversion and git services commented on NIFI-6846:
-------------------------------------------------------
Commit 49b7a7cd6b2645188a732bc57122d1757516e2bd in nifi's branch
refs/heads/master from Mark Payne
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=49b7a7c ]
NIFI-6846: If an Exception is thrown while a Processor is writing to a
FlowFile, but that Content Claim is not yet eligible for destruction, mark it
as a transient claim on the RepositoryRecord so that if it's available when the
FlowFile Repository is checkpointed, then it will be cleaned up then
This closes #3872
> Content Repository may fail to cleanup after Processor throws Exception
> -----------------------------------------------------------------------
>
> Key: NIFI-6846
> URL: https://issues.apache.org/jira/browse/NIFI-6846
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
> Time Spent: 20m
> Remaining Estimate: 0h
>
> If a Processor throws an Exception while writing to a FlowFile, the Content
> Claim that gets created for the FlowFile may not get cleaned up. To
> demonstrate this, create a ConvertRecord processor. For the Record Reader,
> use a CSV Reader. Feed it a CSV file where the content is malformed at some
> point in the middle of the CSV file. Loop the failure relationship back to
> the ConvertRecord processor.
> After the FlowFile has failed several times, the Processor can be stopped. At
> this point, you can determine how many resource claims exist in the Content
> Repository by running the following command:
> `find content_repository -type f | grep -v archive | wc -l`
> This number should be approximately equal to 1 + the number of failures.
> However, after the FlowFile Repository is checkpointed, running the command
> again should yield approximately 1, not 1 + the number of failures. But
> currently, I'm seeing 1 + the number of failures.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
