ProducerFencedException Processing Kafka Stream - apache-kafka

Question

I'm using kafka 1.1.0. A kafka stream consistently throws this exception (albeit with different messages)
WARN o.a.k.s.p.i.RecordCollectorImpl#onCompletion:166 - task [0_0] Error sending record (key KEY value VALUE timestamp TIMESTAMP) to topic OUTPUT_TOPIC due to Producer attempted an operation with an old epoch. Either there is a newer producer with the same transactionalId, or the producer's transaction has been expired by the broker.; No more records will be sent and no more offsets will be recorded for this task.
WARN o.a.k.s.p.i.AssignedStreamsTasks#closeZombieTask:202 - stream-thread [90556797-3a33-4e35-9754-8a63200dc20e-StreamThread-1] stream task 0_0 got migrated to another thread already. Closing it as zombie.
WARN o.a.k.s.p.internals.StreamThread#runLoop:752 - stream-thread [90556797-3a33-4e35-9754-8a63200dc20e-StreamThread-1] Detected a task that got migrated to another thread. This implies that this thread missed a rebalance and dropped out of the consumer group. Trying to rejoin the consumer group now.
org.apache.kafka.streams.errors.TaskMigratedException: StreamsTask taskId: 0_0
ProcessorTopology:
KSTREAM-SOURCE-0000000000:
topics:
[INPUT_TOPIC]
children: [KSTREAM-PEEK-0000000001]
KSTREAM-PEEK-0000000001:
children: [KSTREAM-MAP-0000000002]
KSTREAM-MAP-0000000002:
children: [KSTREAM-SINK-0000000003]
KSTREAM-SINK-0000000003:
topic:
OUTPUT_TOPIC
Partitions [INPUT_TOPIC-0]
at org.apache.kafka.streams.processor.internals.StreamTask.process(StreamTask.java:238)
at org.apache.kafka.streams.processor.internals.AssignedStreamsTasks.process(AssignedStreamsTasks.java:94)
at org.apache.kafka.streams.processor.internals.TaskManager.process(TaskManager.java:411)
at org.apache.kafka.streams.processor.internals.StreamThread.processAndMaybeCommit(StreamThread.java:918)
at org.apache.kafka.streams.processor.internals.StreamThread.runOnce(StreamThread.java:798)
at org.apache.kafka.streams.processor.internals.StreamThread.runLoop(StreamThread.java:750)
at org.apache.kafka.streams.processor.internals.StreamThread.run(StreamThread.java:720)
Caused by: org.apache.kafka.common.errors.ProducerFencedException: task [0_0] Abort sending since producer got fenced with a previous record
I'm not sure what is causing this exception. When I restart application it appears to successfully process a few records before failing with the same exception. Strangely enough, the records are successfully processed several times even though the stream is set to exactly once processing. Here is the stream configuration:
Properties streamProperties = new Properties();
streamProperties.put(StreamsConfig.CACHE_MAX_BYTES_BUFFERING_CONFIG, 0);
streamProperties.put(StreamsConfig.APPLICATION_ID_CONFIG, service.getName());
streamProperties.put(StreamsConfig.PROCESSING_GUARANTEE_CONFIG, "exactly_once");
//Should be DEFAULT_PRODUCTION_EXCEPTION_HANDLER_CLASS_CONFIG - but that field is private.
streamProperties.put("default.production.exception.handler", ErrorHandler.class);
streamProperties.put(StreamsConfig.BOOTSTRAP_SERVERS_CONFIG, brokerUrl);
streamProperties.put(StreamsConfig.REPLICATION_FACTOR_CONFIG, 3);
streamProperties.put(ConsumerConfig.MAX_POLL_RECORDS_CONFIG, 10);
streamProperties.put(KafkaAvroDeserializerConfig.SCHEMA_REGISTRY_URL_CONFIG, schemaRegistryUrl);
streamProperties.put(KafkaAvroDeserializerConfig.SPECIFIC_AVRO_READER_CONFIG, true);
Out of the three servers, only two generate relevant logs when restarting the streams application. Here are logs from the first server:
[2018-05-09 14:42:14,635] INFO [GroupCoordinator 1]: Member INPUT_TOPIC-09dd8ac8-2cd6-4dd1-b963-63ea804c8fcc-StreamThread-1-consumer-3fedb398-91fe-480a-b5ee-1b5879d0956c in group INPUT_TOPIC has failed, removing it from the group (kafka.coordinator.group.GroupCoordinator)
[2018-05-09 14:42:14,636] INFO [GroupCoordinator 1]: Preparing to rebalance group INPUT_TOPIC with old generation 1 (__consumer_offsets-29) (kafka.coordinator.group.GroupCoordinator)
[2018-05-09 14:42:14,636] INFO [GroupCoordinator 1]: Group INPUT_TOPIC with generation 2 is now empty (__consumer_offsets-29) (kafka.coordinator.group.GroupCoordinator)
[2018-05-09 14:42:15,848] INFO [GroupCoordinator 1]: Preparing to rebalance group INPUT_TOPIC with old generation 2 (__consumer_offsets-29) (kafka.coordinator.group.GroupCoordinator)
[2018-05-09 14:42:15,848] INFO [GroupCoordinator 1]: Stabilized group INPUT_TOPIC generation 3 (__consumer_offsets-29) (kafka.coordinator.group.GroupCoordinator)
[2018-05-09 14:42:15,871] INFO [GroupCoordinator 1]: Assignment received from leader for group INPUT_TOPIC for generation 3 (kafka.coordinator.group.GroupCoordinator)
And from the second server:
[2018-05-09 14:42:16,228] INFO [TransactionCoordinator id=0] Initialized transactionalId INPUT_TOPIC-0_0 with producerId 2010 and producer epoch 37 on partition __transaction_state-37 (kafka.coordinator.transaction.TransactionCoordinator)
[2018-05-09 14:44:22,121] INFO [TransactionCoordinator id=0] Completed rollback ongoing transaction of transactionalId: INPUT_TOPIC-0_0 due to timeout (kafka.coordinator.transaction.TransactionCoordinator)
[2018-05-09 14:44:42,263] ERROR [ReplicaManager broker=0] Error processing append operation on partition OUTPUT_TOPIC-0 (kafka.server.ReplicaManager)
org.apache.kafka.common.errors.ProducerFencedException: Producer's epoch is no longer valid. There is probably another producer with a newer epoch. 37 (request epoch), 38 (server epoch)
It appears like the first server sees that the consumer has failed and removes it from the consumer group before it is registered with the second server. Any ideas what could be causing the consumer to fail? Or, any ideas handling this failure gracefully? It's possible that it is this bug, does anyone know of a possible workaround?

Answer 1

I'm not sure what caused the problem, but reducing the max.poll.records to 1 fixed the problem.