Kafka源码分析（三）——Producer：Metadata集群元数据

    // 1.创建Metadata对象
    this.metadata = new Metadata(retryBackoffMs, config.getLong(ProducerConfig.METADATA_MAX_AGE_CONFIG), true, clusterResourceListeners);
    // 2.解析Broker地址
    List<InetSocketAddress> addresses = ClientUtils.parseAndValidateAddresses(config.getList(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG));
    // 3.更新元数据
    this.metadata.update(Cluster.bootstrap(addresses), Collections.<String>emptySet(), time.milliseconds());

首先，KafkaProducer创建了一个MetaData对象；

接着，根据我们自己配置的bootstrap.servers地址，创建一个Cluster对象——Cluster.bootstrap(addresses)；

最后，调用MetaData.update()进行元数据的更新。

1.1 MetaData元数据

我们来看下MetaData到底是个什么东西？从字段定义可以看出来，它内部就是保存了一些Topic的更新策略，同时封装了一个Cluster对象。也就是MetaData只是一个壳，仅仅定义了元数据的更新策略，真正的元数据信息保存在Cluster对象中：

    public final class Metadata {
        public static final long TOPIC_EXPIRY_MS = 5 * 60 * 1000;
        private static final long TOPIC_EXPIRY_NEEDS_UPDATE = -1L;
    
        private final long refreshBackoffMs;
        private final long metadataExpireMs;
        private int version;
        private long lastRefreshMs;
        private long lastSuccessfulRefreshMs;
        private Cluster cluster;
        private boolean needUpdate;
        private final Map<String, Long> topics;
        private final List<Listener> listeners;
        private final ClusterResourceListeners clusterResourceListeners;
        private boolean needMetadataForAllTopics;
        private final boolean topicExpiryEnabled;
    
        public Metadata(long refreshBackoffMs, long metadataExpireMs, boolean topicExpiryEnabled, ClusterResourceListeners clusterResourceListeners) {
            this.refreshBackoffMs = refreshBackoffMs;
            this.metadataExpireMs = metadataExpireMs;
            this.topicExpiryEnabled = topicExpiryEnabled;
            this.lastRefreshMs = 0L;
            this.lastSuccessfulRefreshMs = 0L;
            this.version = 0;
            this.cluster = Cluster.empty();
            this.needUpdate = false;
            this.topics = new HashMap<>();
            this.listeners = new ArrayList<>();
            this.clusterResourceListeners = clusterResourceListeners;
            this.needMetadataForAllTopics = false;
        }
    
        //...
    }

KafkaProducer在初始化时，调用了Metadata的update方法，这个方法是加锁的，也就是说每次只能有一个线程执行更新操作：

    public synchronized void update(Cluster cluster, Set<String> unavailableTopics, long now) {
        this.needUpdate = false;
        this.lastRefreshMs = now;
        this.lastSuccessfulRefreshMs = now;
        this.version += 1;
    
        if (topicExpiryEnabled) {    // 是否允许Topic元数据过期，默认true
            for (Iterator<Map.Entry<String, Long>> it = topics.entrySet().iterator(); it.hasNext(); ) {
                Map.Entry<String, Long> entry = it.next();
                long expir

最低0.47元/天解锁文章