Kafka producer TimeoutException: Expiring 1 record(s)

Question

I am using Kafka with Spring-boot:

Kafka Producer class:

@Service public class MyKafkaProducer {      @Autowired     private KafkaTemplate<String, String> kafkaTemplate;      private static Logger LOGGER = LoggerFactory.getLogger(NotificationDispatcherSender.class);      // Send Message     public void sendMessage(String topicName, String message) throws Exception {         LOGGER.debug("========topic Name===== " + topicName + "=========message=======" + message);         ListenableFuture<SendResult<String, String>> result = kafkaTemplate.send(topicName, message);         result.addCallback(new ListenableFutureCallback<SendResult<String, String>>() {             @Override             public void onSuccess(SendResult<String, String> result) {                 LOGGER.debug("sent message='{}' with offset={}", message, result.getRecordMetadata().offset());             }              @Override             public void onFailure(Throwable ex) {                 LOGGER.error(Constants.PRODUCER_MESSAGE_EXCEPTION.getValue() + " : " + ex.getMessage());             }         });     } } 

Kafka-configuration:

spring.kafka.producer.retries=0 spring.kafka.producer.batch-size=100000 spring.kafka.producer.request.timeout.ms=30000 spring.kafka.producer.linger.ms=10 spring.kafka.producer.acks=0 spring.kafka.producer.buffer-memory=33554432 spring.kafka.producer.max.block.ms=5000 spring.kafka.bootstrap-servers=192.168.1.161:9092,192.168.1.162:9093 

Let's say I have sent 10 times 1000 messages in topic my-test-topic.

8 out of 10 times I am successfully getting all messages in my consumer but sometimes I am getting this below error:

2017-10-05 07:24:11, [ERROR] [my-service - LoggingProducerListener - onError:76] Exception thrown when sending a message with key='null' and payload='{"deviceType":"X","deviceKeys":[{"apiKey":"X-X-o"}],"devices...' to topic my-test-topic

and org.apache.kafka.common.errors.TimeoutException: Expiring 1 record(s) for my-test-topic-4 due to 30024 ms has passed since batch creation plus linger time

Answer 1

There are 3 possibilities:

1. Increase request.timeout.ms - this is the time that Kafka will wait for whole batch to be ready in buffer. So in your case if there are less than 100 000 messages in buffer, timeout will occur. More info here: https://stackoverflow.com/a/34794261/2707179
2. Decrease batch-size - related to previous point, it will send batches more often but they will include fewer messages.
3. Depending on message size, maybe your network cannot catch up with high load? Check if your throughput is not a bottleneck.

Answer 2

1. First clue in the error is 30024 ms has passed - the configuration spring.kafka.producer.request.timeout.ms=30000 is related. This 30 seconds wait is for filling up the buffer on the Producer side.

2. When a message is published, it gets buffered at the Producer's side and it will wait for 30s (see config above) to fill up. spring.kafka.producer.batch-size=100000 means 100KB so if the message ingestion load is low, and the buffer doesn't fill up with more messages to 100KB in 30s, you would expect this message.

3. spring.kafka.producer.linger.ms=10 is used where the ingestion load is high and the producer wants to limit send() calls to Kafka brokers. This is the duration Producer will wait before sending messages to the broker after batch is ready (i.e. after buffer is filled up to batch size of 100KB).

Solution:

• Increase linger.ms to hold messages longer after the batch is ready. If more time is needed to fill the batch, increase request.timeout.ms.
• Another approach: reduce batch-size, or increase request.timeout.ms, or both.