[
https://issues.apache.org/jira/browse/CAMEL-20563?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17827098#comment-17827098
]
Sami Peltola commented on CAMEL-20563:
--------------------------------------
Attached an example project, instructions in the README.md.
Seems that using breakOnFirstError option for Kafka component creates a new
heartbeat thread every time KafkaFetchRecords is set to reconnect, which
happens when there an exception is allowed to propagate back from the route.
In KafkaFetchRecords, the task is set to reconnection state every time an
exception bubbles back from the route:
{code:java}
// when breakOnFirstError we want to unsubscribe from Kafka
if (result != null && result.isBreakOnErrorHit() &&
!this.state.equals(State.PAUSED)) {
LOG.debug("We hit an error ... setting flags to force
reconnect");
// force re-connect
setReconnect(true);
setConnected(false);
}
{code}
This will lead to kafka-client
*kafkaConsumer.getEndpoint().getKafkaClientFactory().getConsumer* being called
again in KafkaFetchRecords.createConsumer(), which in turn creates
ConsumerCoordinator again. The problem is that the heartbeat-thread created for
the previous ConsumerCoordinator is never shutdown.
The new ConsumerCoordinator will have class member heartbeatThread set to null
and the next time ensureActiveGroup is called, a new heartbeat-thread is
created:
{code:java}
private synchronized void startHeartbeatThreadIfNeeded() {
if (heartbeatThread == null) {
heartbeatThread = new HeartbeatThread();
heartbeatThread.start();
}
}
{code}
> camel-kafka - breakOnFirstError causes thread and memory leaks
> --------------------------------------------------------------
>
> Key: CAMEL-20563
> URL: https://issues.apache.org/jira/browse/CAMEL-20563
> Project: Camel
> Issue Type: Bug
> Components: camel-kafka
> Affects Versions: 3.18.5, 3.22.1
> Reporter: Aniket Jadhav
> Priority: Major
> Attachments: KafkaHeartBeatLeakThread.PNG,
> heartbeat-threading-problem.zip
>
>
> This has got fixed in
> [CAMEL-16857|https://issues.apache.org/jira/browse/CAMEL-16857]. But Facing
> same issue with 3.18.5, is it reintroduced at some point?
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
