[KAFKA-5242] add max_number _of_retries to exponential backoff strategy

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Won't Do
• Affects Version/s: 0.10.2.0
• Fix Version/s: 0.10.2.2, 0.11.0.1
• Component/s: streams
• Assignee: Matthias J. Sax
• Reporter: Lukas Gemela
• Created: 15/May/17 14:44
• Updated: 17/Aug/17 08:57
• Resolved: 17/Aug/17 08:57

Description

From time to time, during relabance we are getting a lot of exceptions saying

org.apache.kafka.streams.errors.LockException: task [0_0] Failed to lock the state directory: /app/db/clio/0_0
	at org.apache.kafka.streams.processor.internals.ProcessorStateManager.<init>(ProcessorStateManager.java:102) ~[kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.AbstractTask.<init>(AbstractTask.java:73) ~[kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamTask.<init>(StreamTask.java:108) ~[kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread.createStreamTask(StreamThread.java:834) ~[kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread$TaskCreator.createTask(StreamThread.java:1207) ~[kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread$AbstractTaskCreator.retryWithBackoff(StreamThread.java:1180) [kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread.addStreamTasks(StreamThread.java:937) [kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread.access$500(StreamThread.java:69) [kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread$1.onPartitionsAssigned(StreamThread.java:236) [kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.onJoinComplete(ConsumerCoordinator.java:255) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.internals.AbstractCoordinator.joinGroupIfNeeded(AbstractCoordinator.java:339) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.internals.AbstractCoordinator.ensureActiveGroup(AbstractCoordinator.java:303) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.internals.ConsumerCoordinator.poll(ConsumerCoordinator.java:286) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.KafkaConsumer.pollOnce(KafkaConsumer.java:1030) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:995) [kafka-clients-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread.runLoop(StreamThread.java:582) [kafka-streams-0.10.2.0.jar!/:?]
	at org.apache.kafka.streams.processor.internals.StreamThread.run(StreamThread.java:368) [kafka-streams-0.10.2.0.jar!/:?]

(see attached logfile)

It was actually problem on our side - we ran startStreams() twice and therefore we had two threads touching the same folder structure.

But what I've noticed, the backoff strategy in StreamThread$AbstractTaskCreator.retryWithBackoff can run endlessly - after 20 iterations it takes 6hours until the next attempt to start a task.
I've noticed latest code contains check for rebalanceTimeoutMs, but that still does not solve the problem especially in case MAX_POLL_INTERVAL_MS_CONFIG is set to Integer.MAX_INT. at this stage kafka streams just hangs up indefinitely.

I would personally make that backoffstrategy a bit more configurable with a number of retries that if it exceed a configured value it propagates the exception as any other exception to custom client exception handler.
(I can provide a patch)

Issue Links

• relates to: KAFKA-5152 Kafka Streams keeps restoring state after shutdown is initiated during startup