Kevin Doran created NIFI-15801:
----------------------------------
Summary:
StandardVersionedComponentSynchronizer.synchronizeProcessors fails when
processors are running
Key: NIFI-15801
URL: https://issues.apache.org/jira/browse/NIFI-15801
Project: Apache NiFi
Issue Type: Task
Components: Core Framework
Reporter: Kevin Doran
Assignee: Kevin Doran
`StandardVersionedComponentSynchronizer.synchronizeProcessors()` calls
`updateProcessor()` without first stopping the processor. `updateProcessor()`
begins with `processor.setAnnotationData(...)`, and
`StandardProcessorNode.setAnnotationData()` throws `IllegalStateException` if
the processor is running.
This was never triggered before the introduction of Connectors, because the
NiFi REST API `FlowUpdateResource.updateFlow()` explicitly stops all affected
processors and disables affected controller services *before* calling
`StandardProcessGroup.updateFlow()` -> `synchronizeFlow()` -> `synchronize()`
-> `synchronizeProcessors()`.
The Connector framework's `applyUpdate` path, introduced in NIFI-15258, calls
`StandardProcessGroup.updateFlow()` directly without stopping processors first,
exposing this gap.
The single-processor synchronization path (`synchronize(ProcessorNode,
VersionedProcessor, ...)`) already correctly calls `stopOrTerminate()` before
`updateProcessor()`. The group-level `synchronizeProcessors()` was relying on
the caller to have stopped processors, which was a valid assumption for the
REST path.
**Steps to Reproduce:**
1. Start a Connector
2. While the connector is running, change a configuration property that affects
an existing processor in the flow.
3. Call `applyUpdate()`.
**Expected:** The flow is updated successfully; affected processors are
stopped, updated, and restarted.
**Actual:** `FlowUpdateException` wrapping `IllegalStateException: Cannot set
AnnotationData on CaptureChangePostgreSQL[...] while processor is running`.
**Fix:** Stop running processors in `synchronizeProcessors()` before calling
`updateProcessor()`, matching the pattern used in `synchronize(ProcessorNode,
...)` and `synchronizeProcessGroupSettings()`. Track which processors were
stopped and restart them in a `finally` block to ensure they are restarted even
if the update fails.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
