Apache Kafka & Stream Processing Fundamentals · 课时

从 Kafka 消费消息

学习构建消费者，从 Kafka 主题读取并处理数据流，同时处理偏移量和消费者组

从 Kafka 消费消息 是 CoddyKit 上的免费 Apache Kafka & Stream Processing Fundamentals 课时。这是第 2 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 Apache Kafka & Stream Processing Fundamentals 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Apache Kafka & Stream Processing Fundamentals 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

What are Kafka Consumers?

Kafka Consumers are applications that read data from Kafka topics. They subscribe to one or more topics and process the messages as they arrive.

Consumers are the 'listeners' in your Kafka ecosystem.
They pull data from Kafka brokers, rather than having data pushed to them.
Essential for building real-time data pipelines and analytics systems.

Working with Consumer Groups

Consumers often work together in a consumer group. This is a collection of consumer instances that share a common group.id.

Scalability: Messages from a topic's partitions are distributed among consumers in the same group, allowing parallel processing.
Fault Tolerance: If a consumer instance fails, its assigned partitions are automatically reassigned to other active consumers in the group.
Each message in a partition is delivered to only one consumer within a group.

How Consumers Read Data (Polling)

Kafka consumers don't just 'receive' messages. Instead, they actively poll Kafka brokers for new data.

The consumer repeatedly asks Kafka, "Do you have any new records for me?"
This polling mechanism gives the consumer control over its processing rate.
If no new records are available, the poll() method will simply return an empty list after waiting for a specified duration.

Basic Java Consumer Setup

To create a Kafka consumer in Java, you use the KafkaConsumer class. First, you need to set up some basic properties like the Kafka server address and deserializers.

Here's the start of a simple consumer program:

import org.apache.kafka.clients.consumer.KafkaConsumer;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.common.serialization.StringDeserializer;

import java.util.Properties;
import java.util.Collections;

public class SimpleConsumerSetup {
    public static void main(String[] args) {
        // 1. Define consumer properties
        Properties props = new Properties();
        props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        props.put(ConsumerConfig.GROUP_ID_CONFIG, "my-first-group");
        props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());
        props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());

        // 2. Create the consumer
        KafkaConsumer<String, String> consumer = new KafkaConsumer<>(props);

        // 3. Subscribe to a topic
        consumer.subscribe(Collections.singletonList("my-topic"));

        // Consumer loop and closing will go here
        System.out.println("Consumer setup complete. Subscribed to 'my-topic'.");
        consumer.close();
    }
}

Essential Consumer Properties

Let's break down the key properties we set for our consumer:

BOOTSTRAP_SERVERS_CONFIG: A list of Kafka broker addresses (e.g., localhost:9092) that the consumer will connect to.
GROUP_ID_CONFIG: A unique identifier for the consumer group this consumer belongs to (e.g., my-first-group).
KEY_DESERIALIZER_CLASS_CONFIG: Specifies how to convert message keys from bytes (received from Kafka) into Java objects (e.g., StringDeserializer for text keys).
VALUE_DESERIALIZER_CLASS_CONFIG: Specifies how to convert message values from bytes into Java objects (e.g., StringDeserializer for text values).

The Consumer Loop: Polling for Records

Consumers continuously poll Kafka for new data. This is typically done in an infinite loop. The poll(Duration timeout) method returns a batch of ConsumerRecords or an empty list if no new records are available within the specified timeout.

import org.apache.kafka.clients.consumer.KafkaConsumer;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.clients.consumer.ConsumerRecords;
import org.apache.kafka.common.serialization.StringDeserializer;

import java.time.Duration;
import java.util.Properties;
import java.util.Collections;

public class PollingConsumer {
    public static void main(String[] args) {
        Properties props = new Properties();
        props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        props.put(ConsumerConfig.GROUP_ID_CONFIG, "my-polling-group");
        props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());
        props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());

        KafkaConsumer<String, String> consumer = new KafkaConsumer<>(props);
        consumer.subscribe(Collections.singletonList("my-topic"));

        try {
            while (true) { // Infinite loop to continuously poll
                ConsumerRecords<String, String> records = consumer.poll(Duration.ofMillis(100));
                if (records.isEmpty()) {
                    System.out.println("No records received. Waiting...");
                } else {
                    System.out.println("Received " + records.count() + " records.");
                    // Processing of records would happen here
                }
            }
        } finally {
            consumer.close();
            System.out.println("Consumer closed.");
        }
    }
}

Processing Received Messages

Once you receive ConsumerRecords from poll(), you can iterate over them to process each message individually. Each ConsumerRecord contains the topic, partition, offset, key, and value of the message.

This is where your application's specific logic for handling the data lives.

import org.apache.kafka.clients.consumer.KafkaConsumer;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.clients.consumer.ConsumerRecords;
import org.apache.kafka.clients.consumer.ConsumerRecord;
import org.apache.kafka.common.serialization.StringDeserializer;

import java.time.Duration;
import java.util.Properties;
import java.util.Collections;

public class ProcessingConsumer {
    public static void main(String[] args) {
        Properties props = new Properties();
        props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        props.put(ConsumerConfig.GROUP_ID_CONFIG, "my-processing-group");
        props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());
        props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());

        KafkaConsumer<String, String> consumer = new KafkaConsumer<>(props);
        consumer.subscribe(Collections.singletonList("my-topic"));

        try {
            while (true) {
                ConsumerRecords<String, String> records = consumer.poll(Duration.ofMillis(100));
                for (ConsumerRecord<String, String> record : records) {
                    System.out.printf("Topic: %s, Partition: %d, Offset: %d, Key: %s, Value: %s%n",
                        record.topic(), record.partition(), record.offset(), record.key(), record.value());
                    // Add your custom processing logic here, e.g., store in database, perform analytics
                }
            }
        } finally {
            consumer.close();
            System.out.println("Consumer closed.");
        }
    }
}

Tracking Progress with Offsets

An offset is a unique, sequential ID assigned to each message within a Kafka topic partition. It acts like a pointer, indicating the position of a message.

Consumers use offsets to track which messages they have already processed.
Kafka stores the latest committed offset for each consumer group and partition.
This ensures that if a consumer restarts or a partition is reassigned, it knows exactly where to resume reading to avoid reprocessing or missing messages.

Offsets are crucial for reliable message processing.

Automatic Offset Committing

By default, Kafka consumers are configured to auto-commit offsets periodically. This is convenient for simple applications but might lead to message loss or duplicates in certain failure scenarios.

Set enable.auto.commit to true (this is the default setting).
The auto.commit.interval.ms property determines how often offsets are committed (e.g., every 5 seconds).
This approach is generally suitable for 'at-least-once' processing where some duplicates are acceptable.

Manual Offset Control

For more precise control over message processing and stronger guarantees (like 'exactly-once' processing), you can manually commit offsets after your application has successfully processed a batch of messages.

Set enable.auto.commit to false to disable automatic commits.
Use consumer.commitSync(): This method blocks until the commit operation is successful.
Use consumer.commitAsync(): This method is non-blocking and faster, but requires a callback function for handling potential errors.

Manual commits ensure that records are processed *before* their offsets are marked as 'done', preventing data loss on consumer failure.

import org.apache.kafka.clients.consumer.KafkaConsumer;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.clients.consumer.ConsumerRecords;
import org.apache.kafka.clients.consumer.ConsumerRecord;
import org.apache.kafka.common.serialization.StringDeserializer;

import java.time.Duration;
import java.util.Properties;
import java.util.Collections;

public class ManualCommitConsumer {
    public static void main(String[] args) {
        Properties props = new Properties();
        props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        props.put(ConsumerConfig.GROUP_ID_CONFIG, "my-manual-commit-group");
        props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());
        props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());
        props.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, "false"); // Disable auto-commit

        KafkaConsumer<String, String> consumer = new KafkaConsumer<>(props);
        consumer.subscribe(Collections.singletonList("my-topic"));

        try {
            while (true) {
                ConsumerRecords<String, String> records = consumer.poll(Duration.ofMillis(100));
                for (ConsumerRecord<String, String> record : records) {
                    System.out.printf("Processed: Partition %d, Offset %d, Value: %s%n",
                        record.partition(), record.offset(), record.value());
                    // Simulate processing work that must complete before committing
                }
                if (!records.isEmpty()) {
                    consumer.commitSync(); // Commit offsets after processing the batch
                    System.out.println("Offsets committed manually.");
                }
            }
        } finally {
            consumer.close();
            System.out.println("Consumer closed.");
        }
    }
}

Check Your Understanding

Consider a Kafka topic with 3 partitions. A consumer group with 2 consumer instances is subscribed to this topic. If one consumer instance fails, what happens to the partitions it was processing, and how does Kafka ensure no data is lost?

Lesson Recap: Consuming Data

Great job! In this lesson, you learned how Kafka consumers read data:

Consumers subscribe to topics and actively poll Kafka for new messages.
Consumer groups enable scalable and fault-tolerant message processing across multiple consumer instances.
Offsets are sequential IDs that track a consumer's progress within a topic partition.
You can choose between automatic (simpler, default) or manual (more control, better guarantees) offset committing for reliability.

Next, we'll dive deeper into partitions and offsets to understand how Kafka achieves high throughput and message ordering.

常见问题解答

「从 Kafka 消费消息」课时是免费的吗？

是的 — 「从 Kafka 消费消息」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Apache Kafka & Stream Processing Fundamentals 课程的其余内容，请升级到 CoddyKit PRO。 Apache Kafka & Stream Processing Fundamentals 课程共包含 4 节课。

「从 Kafka 消费消息」这节课中我会学到什么？

学习构建消费者，从 Kafka 主题读取并处理数据流，同时处理偏移量和消费者组你通过在浏览器中直接运行的动手代码来练习 Apache Kafka & Stream Processing Fundamentals，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 Apache Kafka & Stream Processing Fundamentals 需要有经验吗？

无需任何先前经验。CoddyKit 上的 Apache Kafka & Stream Processing Fundamentals 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 2 节课，共 4 节。

「从 Kafka 消费消息」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 Apache Kafka & Stream Processing Fundamentals 课中编写并运行代码吗？

能。每节 Apache Kafka & Stream Processing Fundamentals 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 Apache Kafka & Stream Processing Fundamentals