[
https://issues.apache.org/jira/browse/NIFI-1959?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15313279#comment-15313279
]
ASF GitHub Bot commented on NIFI-1959:
--------------------------------------
Github user pvillard31 commented on the issue:
https://github.com/apache/nifi/pull/490
It remains an issue I have not been able to solve at this point: the case
where the new file contains more data than the tailed file before being moved.
In this case we ends up in the same situation: the new data in not ingested and
the processor is not yielded. This situation should not happen really often...
but still... if someone has an idea.
To reproduce: tail on /tmp/test.log
echo "test" >> /tmp/test.log
rm /tmp/test.log
echo "testtest" >> /tmp/test.log
In this case, both timestamp and size cannot be used to detect a
pseudo-rollover and we are stuck with the previous state and reader. Not sure
to see an easy solution.
> TailFile not ingesting data when tailed file is moved with no rolling pattern
> -----------------------------------------------------------------------------
>
> Key: NIFI-1959
> URL: https://issues.apache.org/jira/browse/NIFI-1959
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
> Affects Versions: 0.6.1
> Reporter: Pierre Villard
> Assignee: Pierre Villard
> Priority: Minor
>
> In case "Rolling Filename Pattern" is not set by user and if the tailed file
> is moved, then the new file will not be tailed.
> Besides, in such case, the processor will be endlessly triggered without
> ingesting data: it creates a lot of tasks and consumes CPU. The reason is it
> never goes in if statement L448.
> A solution is to look at size() and lastUpdated() of the tailed file to
> detect a "rollover". However it won't allow the processor to ingest the
> potential data added in the tailed file just before being moved.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
