[
https://issues.apache.org/jira/browse/NIFI-15885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18077383#comment-18077383
]
ASF subversion and git services commented on NIFI-15885:
--------------------------------------------------------
Commit 0574c82f0247031a5af734328885f8e7530a7c53 in nifi's branch
refs/heads/main from Mark Payne
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=0574c82f024 ]
NIFI-15885: Addressed issues that cause Offload to get stuck: when a … (#11184)
* NIFI-15885: Addressed issues that cause Offload to get stuck: when a
processor extends AbstractSessionFactoryProcessor and stores a reference to a
created ProcessSession but not its factory, garbage collection could cause
framework to be unable to rollback sessions. When a processor is terminated,
ensure we rollback retained sessions even if all threads are completed. Fixed
related bug that caused background thread never to complete when waiting for
processor thread count to reach 1 and terminate set it to 0.
* NIFI-15885: Fix to unit test
* NIFI-15885: Allow StandardProcessSession.migrate to unwrap delegating Session
wrappers so FactoryRetainingProcessSession (used to keep an
ActiveProcessSessionFactory reachable for offload/terminate) is recognized as a
StandardProcessSession target. Adds a framework-internal
DelegatingProcessSession contract implemented by the wrapper, and an IT
covering the migrate-to-wrapper path that reproduces the MergeRecord CI failure.
> Occasionally NiFi node fails to complete offload / decommission
> ---------------------------------------------------------------
>
> Key: NIFI-15885
> URL: https://issues.apache.org/jira/browse/NIFI-15885
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> Occasionally when a NiFi node is decommissioned, it transitions to
> 'OFFLOADING' state and never completes. It just gets 'stuck' there, showing
> that it's still waiting for a few FlowFiles to complete offload. This implies
> that the stop/terminate is not always properly rolling back the session that
> have pulled data from queues.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
