Mark Payne created NIFI-14663:
---------------------------------
Summary: Threading bug can allow Processor to report that it is
stopped before it truly is
Key: NIFI-14663
URL: https://issues.apache.org/jira/browse/NIFI-14663
Project: Apache NiFi
Issue Type: Bug
Components: Core Framework
Reporter: Mark Payne
Assignee: Mark Payne
Attachments: image-2025-06-14-11-16-22-089.png
In order to determine if a Processor is fully stopped, we must do 2 things: We
must check that the state is STOPPED and we must also check that there are no
active threads.
However, when ProcessorNode.stop() is called, it transitions the state from
RUNNING to STOPPING (STOPPING is an "internal-only" state that is never exposed
outside of the ProcessorNode class) and then immediately increment the Active
Thread Count (unless there are no lifecycle methods to call).
This can result in a Processor being stopped, which transitions its internal
state to STOPPING and then ProcessorNode.isRunning() will return false.
Immediately afterward, the Active Thread count can be incremented to 1 in order
to call @OnUnscheduled methods, and ProcessorNode.isRunning() will return true
- even though it hasn't been restated.
This has been observed in a failure of
FlowSynchronizationTestIT.testAddControllerServiceReferencingExistingService()
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
