GlobalKTable - StreamsException：遇到与任何全局状态存储无关的主题分区

Question

我正在尝试使用 Kafka-streams 从流中创建 GlobalKTable，并在调用streams.start()时出现异常：

org.apache.kafka.streams.errors.StreamsException：遇到与任何全局状态存储无关的主题分区

我的代码是：

private final KafkaStreams streams;
private final StoreQueryParameters<ReadOnlyKeyValueStore<LocalDate, String>> bankHolidayTypesSqp = StoreQueryParameters.fromNameAndType("bank_holiday_type_store"
            ,QueryableStoreTypes.<LocalDate, String>keyValueStore());
private final ReadOnlyKeyValueStore<LocalDate, String> localBankHolidayTypeStore;

private void instantiateKafka()
{
        // configure Kafka
        
        StreamsBuilder builder = new StreamsBuilder();
       
       // CustomSerializableSerde is just a generic serializer that uses standard java Base64 encoding on any object that implements Serializable - it works in a dummy application I've tested, so I don't think it's the problem
        addGlobalTableToStreamsBuilder(builder, bankHolidayTypeTopic,"bank_holiday_type_store", new CustomSerializableSerde<LocalDate>(),Serdes.String());
        
        streams = createStreams("localhost:9092", "C:\\Kafka\\tmp\\kafka-streams-global-tables",MyClass.class.getName(),builder);
        streams.start(); // hangs until the global table is built
}

public static <Tk extends Serializable,Tv extends Serializable> StreamsBuilder addGlobalTableToStreamsBuilder(StreamsBuilder builder, String topic
            , String store, Serde<Tk> keySerializer, Serde<Tv> valueSerializer)
    {
        builder.globalTable(topic, Materialized.<Tk, Tv, KeyValueStore<Bytes, byte[]>>as(store)
                .withKeySerde(keySerializer)
                .withValueSerde(valueSerializer));
        return builder;
    }

public static KafkaStreams createStreams(final String bootstrapServers, final String stateDir, String clientID, StreamsBuilder finishedBuilder) 
     {
        final Properties streamsConfiguration = new Properties();
        // Give the Streams application a unique name.  The name must be unique in the Kafka cluster against which the application is run.
        streamsConfiguration.put(StreamsConfig.APPLICATION_ID_CONFIG, "applicationName");
        streamsConfiguration.put(StreamsConfig.CLIENT_ID_CONFIG, clientID);
        // Where to find Kafka broker(s).
        streamsConfiguration.put(StreamsConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers);
        streamsConfiguration.put(StreamsConfig.STATE_DIR_CONFIG, stateDir);
        // Set to earliest so we don't miss any data that arrived in the topics before the process started
        streamsConfiguration.put(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "earliest");
                
         return new KafkaStreams(finishedBuilder.build(), streamsConfiguration);
     }
    

制作人：

Producer<LocalDate,String> bankHolidayTypeProducer = MyClass<LocalDate,String>createProducer("localhost:9092", BankHolidayData.class.getName()
                    , CustomSerializer.class.getName(), StringSerializer.class.getName());

//...

HashMap<LocalDate, String> bankHolidaysData = populateBankHolidayMap();

for (LocalDate bhDay : bankHolidaysData.keySet())
            {
                bankHolidayTypeProducer.send(new ProducerRecord<>(bankHolidayTypeTopic, bhDay, bankHolidaysData.get(bhDay)));
            }

public static <Tk extends Serializable, Tv extends Serializable> Producer<Tk,Tv> createProducer(String bootstrapServers
            , String clientID, String keySerializerClassName, String valueSerializerClassName) 
    {
        Properties props = new Properties();
        props.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers);
        props.put(ProducerConfig.CLIENT_ID_CONFIG, clientID);
        props.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, keySerializerClassName);
        props.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, valueSerializerClassName);
        return new KafkaProducer<>(props);
    }

我的主题由生产者在首次生成时自动创建，并且在 GlobalKTable 尝试从中读取时将始终存在。 这是问题吗？ 在设置主题时我需要做些什么来告诉 Kafka 它将被 Streams GlobalKTable 使用吗？

Answer 1

主题的结构（显然）发生了一些变化，这意味着需要重置 Streams。 为此，您可以使用应用程序 Conduktor，或在http://docs.confluent.io/current/streams/developer-guide.html#application-reset-tool 中找到的重置工具。

Answer 2

如果这对某人有帮助，那么如果您在 GlobalKTable 的消费者端更新任何模式，并且在 Kafka 端的模式发生变化时，也会发生这种情况。 所以对我来说修复的只是删除我的本地状态存储文件夹，通常存在于您的项目根目录中。

Answer 3

如果您正在运行本地实例/或者您有能力，那么

• 停止经纪人
• 停止动物园管理员
• 删除环境变量 java.io.tmpdir 引用的父文件夹下的 kafka-streams 文件夹。

您可以在 Broker 日志中获取真实的文件夹名称，搜索“INFO Client environment:java.io.tmpdir”