Chris Egerton created KAFKA-10873:
-------------------------------------
Summary: Inaccurate "Ignore stop request for unowned connector"
log messages
Key: KAFKA-10873
URL: https://issues.apache.org/jira/browse/KAFKA-10873
Project: Kafka
Issue Type: Bug
Components: KafkaConnect
Reporter: Chris Egerton
If a connector fails during startup, it will never be added to the worker's
internal list of running connectors (see
[https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L298|https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L298).]),
and the same happens for tasks (see
[https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L570]).
This leads to the following {{WARN}}-level log messages when that
connector/task is scheduled to be stopped by the worker:
* [Ignoring stop request for unowned connector
<name>|https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L390]
* [Ignoring await stop request for non-present connector
<name>|https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L415]
* [Ignoring stop request for unowned task
<name>|https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L832]
* [Ignoring await stop request for non-present task
<name>|https://github.com/apache/kafka/blob/87260a33b01590d0f73577840422e24fa589bed0/connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java#L862]
If the connector/task failed on startup, there should already be log messages
detailing the failure and its cause; there is no need to emit warning messages
about not stopping that connector when it is scheduled for shutdown. Even
worse, emitting these messages may cause users to believe that their cluster is
experiencing a rebalancing bug that is somehow causing connectors/tasks that
are not assigned to a worker to be revoked from it.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
