Kafka Consumer Error Handling

Robust error handling ensures consumers can recover from failures without data loss or infinite retry loops. This guide covers error types, handling strategies, and resilience patterns.

---

Error Categories

Error Classification

Category	Description	Examples
Retriable	Temporary failures	Network timeout, leader election
Non-retriable	Permanent failures	Authorization, unknown topic
Poison pill	Bad message data	Deserialization failure, schema mismatch
Processing	Application logic failure	Business rule violation, external service down

Error Handling Strategy

---

Deserialization Errors

Problem

Invalid messages cause deserialization failures:

// Without error handling, bad message crashes consumer
ConsumerRecords<String, Order> records = consumer.poll(Duration.ofMillis(100));
// Throws SerializationException on corrupt message

ErrorHandlingDeserializer

Kafka provides a wrapper deserializer that captures errors:

Properties props = new Properties();
props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, ErrorHandlingDeserializer.class);
props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, ErrorHandlingDeserializer.class);
props.put(ErrorHandlingDeserializer.KEY_DESERIALIZER_CLASS, StringDeserializer.class);
props.put(ErrorHandlingDeserializer.VALUE_DESERIALIZER_CLASS, JsonDeserializer.class);

Handling Poison Pills

while (running) {
    ConsumerRecords<String, Order> records = consumer.poll(Duration.ofMillis(100));

    for (ConsumerRecord<String, Order> record : records) {
        // Check for deserialization failure
        if (record.value() == null) {
            // With ErrorHandlingDeserializer, failures are captured in headers
            byte[] rawValue = getRawValue(record);  // Access raw bytes if needed
            log.error("Deserialization failed at offset {}: {}",
                record.offset(), new String(rawValue));

            // Send to dead letter queue
            sendToDeadLetterQueue(record);
            continue;
        }

        // Process valid message
        processOrder(record.value());
    }

    consumer.commitSync();
}

Custom Deserializer with Error Handling

public class SafeJsonDeserializer<T> implements Deserializer<T> {
    private final ObjectMapper mapper = new ObjectMapper();
    private final Class<T> targetType;

    @Override
    public T deserialize(String topic, byte[] data) {
        if (data == null) return null;

        try {
            return mapper.readValue(data, targetType);
        } catch (IOException e) {
            // Return null instead of throwing
            log.error("Failed to deserialize message: {}", new String(data), e);
            return null;
        }
    }
}

---

Dead Letter Queue

DLQ Pattern

Route failed messages to a dedicated topic for later analysis:

DLQ Implementation

public class DeadLetterQueueHandler {
    private final Producer<String, byte[]> dlqProducer;
    private final String dlqTopicSuffix = ".dlq";

    public void sendToDeadLetterQueue(ConsumerRecord<?, ?> record, Exception error) {
        String dlqTopic = record.topic() + dlqTopicSuffix;

        ProducerRecord<String, byte[]> dlqRecord = new ProducerRecord<>(
            dlqTopic,
            record.key() != null ? record.key().toString() : null,
            serializeOriginalValue(record)
        );

        // Add metadata headers
        dlqRecord.headers()
            .add("dlq.original.topic", record.topic().getBytes())
            .add("dlq.original.partition", Integer.toString(record.partition()).getBytes())
            .add("dlq.original.offset", Long.toString(record.offset()).getBytes())
            .add("dlq.error.message", error.getMessage().getBytes())
            .add("dlq.error.class", error.getClass().getName().getBytes())
            .add("dlq.timestamp", Instant.now().toString().getBytes());

        dlqProducer.send(dlqRecord, (metadata, exception) -> {
            if (exception != null) {
                log.error("Failed to send to DLQ: {}", record, exception);
            }
        });
    }
}

DLQ Topic Configuration

# Create DLQ topic with appropriate settings
cleanup.policy=delete
retention.ms=2592000000  # 30 days

---

Retry Strategies

In-Memory Retry

Simple retry with backoff:

public class RetryingConsumer {
    private final int maxRetries = 3;
    private final long initialBackoffMs = 100;
    private final double backoffMultiplier = 2.0;

    public void processWithRetry(ConsumerRecord<String, String> record) {
        int attempt = 0;
        long backoff = initialBackoffMs;

        while (attempt < maxRetries) {
            try {
                process(record);
                return;  // Success
            } catch (RetriableException e) {
                attempt++;
                log.warn("Attempt {} failed for offset {}: {}",
                    attempt, record.offset(), e.getMessage());

                if (attempt < maxRetries) {
                    Thread.sleep(backoff);
                    backoff = (long) (backoff * backoffMultiplier);
                }
            }
        }

        // Max retries exceeded
        dlqHandler.sendToDeadLetterQueue(record, lastException);
    }
}

Retry Topic Pattern

Use dedicated retry topics for delayed reprocessing:

Implementation

public class RetryTopicConsumer {
    private final Map<String, RetryConfig> retryConfigs = Map.of(
        "orders", new RetryConfig(3, List.of(
            new RetryLevel("orders.retry.1", Duration.ofMinutes(1)),
            new RetryLevel("orders.retry.2", Duration.ofMinutes(10)),
            new RetryLevel("orders.retry.3", Duration.ofMinutes(60))
        ))
    );

    @KafkaListener(topics = {"orders", "orders.retry.*"})
    public void consume(ConsumerRecord<String, String> record) {
        try {
            process(record);
        } catch (RetriableException e) {
            int currentRetry = getRetryLevel(record.topic());
            RetryConfig config = retryConfigs.get(getBaseTopic(record.topic()));

            if (currentRetry < config.maxRetries()) {
                RetryLevel nextLevel = config.levels().get(currentRetry);
                sendToRetryTopic(record, nextLevel.topic(), nextLevel.delay());
            } else {
                sendToDeadLetterQueue(record, e);
            }
        }
    }

    private void sendToRetryTopic(ConsumerRecord<String, String> record,
                                   String retryTopic, Duration delay) {
        ProducerRecord<String, String> retryRecord = new ProducerRecord<>(
            retryTopic,
            record.key(),
            record.value()
        );

        // Add retry metadata
        retryRecord.headers()
            .add("retry.original.topic", record.topic().getBytes())
            .add("retry.timestamp", Instant.now().toString().getBytes())
            .add("retry.delay.ms", Long.toString(delay.toMillis()).getBytes());

        // Kafka has no native per-message delay. Delay can be implemented via:
        // 1. Topic-level delay (requires Kafka feature or custom implementation)
        // 2. Consumer-side delay before processing
        // 3. External scheduler
        producer.send(retryRecord);
    }
}

---

Error Types and Handling

Retriable Errors

Error	Handling
TimeoutException	Retry with backoff
NotLeaderOrFollowerException	Automatic retry by client
DisconnectException	Reconnect and retry
RetriableException	Application-defined retry

try {
    process(record);
} catch (TimeoutException | RetriableException e) {
    // Retry with backoff
    retryWithBackoff(record, e);
} catch (NonRetriableException e) {
    // Send to DLQ immediately
    sendToDeadLetterQueue(record, e);
}

Non-Retriable Errors

Error	Handling
AuthorizationException	Alert and fail
InvalidTopicException	Configuration error, fix and restart
SerializationException	DLQ (poison pill); without ErrorHandlingDeserializer, this fails poll() before a record is available
RecordTooLargeException	DLQ or skip

Commit Errors

try {
    consumer.commitSync();
} catch (CommitFailedException e) {
    // Partition was reassigned during processing
    // Work will be reprocessed by new owner
    log.warn("Commit failed, partition reassigned: {}", e.getMessage());
} catch (TimeoutException e) {
    // Retry commit
    retryCommit();
}

---

Circuit Breaker Pattern

Prevent cascading failures when external dependencies are down:

public class CircuitBreakerConsumer {
    private final CircuitBreaker circuitBreaker = CircuitBreaker.ofDefaults("external-service");

    public void process(ConsumerRecord<String, String> record) {
        Supplier<String> decoratedSupplier = CircuitBreaker
            .decorateSupplier(circuitBreaker, () -> callExternalService(record));

        Try.ofSupplier(decoratedSupplier)
            .recover(CallNotPermittedException.class, e -> {
                // Circuit is open - pause consumer
                pauseConsumer();
                scheduleResume();
                throw e;
            })
            .recover(throwable -> {
                // Other failures - retry or DLQ
                handleFailure(record, throwable);
                return null;
            });
    }

    private void pauseConsumer() {
        consumer.pause(consumer.assignment());
    }

    private void scheduleResume() {
        scheduler.schedule(() -> {
            consumer.resume(consumer.assignment());
        }, 30, TimeUnit.SECONDS);
    }
}

---

Error Monitoring

Metrics

public class ConsumerMetrics {
    private final MeterRegistry registry;

    public void recordProcessingSuccess(String topic) {
        registry.counter("consumer.processing.success", "topic", topic).increment();
    }

    public void recordProcessingError(String topic, String errorType) {
        registry.counter("consumer.processing.error",
            "topic", topic,
            "error_type", errorType
        ).increment();
    }

    public void recordDlqSent(String topic) {
        registry.counter("consumer.dlq.sent", "topic", topic).increment();
    }

    public void recordRetry(String topic, int attemptNumber) {
        registry.counter("consumer.retry",
            "topic", topic,
            "attempt", String.valueOf(attemptNumber)
        ).increment();
    }
}

Alerting

Metric	Alert Condition
DLQ rate	> 1% of messages
Retry rate	> 5% of messages
Processing errors	> 10 per minute
Circuit breaker open	Any open circuit

---

Best Practices

Error Handling Strategy

Practice	Recommendation
Classify errors	Distinguish retriable vs non-retriable
Implement DLQ	Never lose messages
Limit retries	Prevent infinite loops
Use backoff	Avoid overwhelming dependencies

Monitoring

Practice	Recommendation
Track error rates	Alert on anomalies
Log error details	Include offset, partition, error type
Monitor DLQ	Process DLQ messages regularly
Track retry rates	High retry rate indicates problems

Recovery

Practice	Recommendation
DLQ reprocessing	Build tooling to replay DLQ
Error analysis	Review DLQ patterns regularly
Automated recovery	Auto-retry DLQ on fixes

---

Common Patterns

Skip and Log

For non-critical messages:

try {
    process(record);
} catch (Exception e) {
    log.error("Skipping message at offset {}: {}", record.offset(), e.getMessage());
    // Continue processing - message is lost
}

Fail Fast

For critical processing:

try {
    process(record);
} catch (Exception e) {
    log.error("Critical error, stopping consumer: {}", e.getMessage());
    running.set(false);
    throw e;
}

Graceful Degradation

For high-availability requirements:

try {
    processWithFullEnrichment(record);
} catch (EnrichmentServiceException e) {
    // Fall back to basic processing
    processBasic(record);
}

---

Related Documentation

• Consumer Guide - Consumer patterns
• Offset Management - Commit handling
• Anti-Patterns - Common mistakes
• Delivery Semantics - Processing guarantees