CAP理论之CP模型ZK、AP模型Eureka

CAP：

  C：一致性>Consistency;

      取舍：(强一致性、单调一致性、会话一致性、最终一致性、弱一致性)

  A：可用性>Availability;

  P：分区容错性>Partition tolerance;

1、ZK(文件系统+监听机制watcher<观察者模式>)：

  配置JAVA环境：java -version;

  下载解压zookeeper：

  wget http://mirror.bit.edu.cn/apache/zookeeper/stable/zookeeper-3.4.12.tar.gz;

  tar -zxvf zookeeper-3.4.12.tar.gz; *.sh start;

  重命名配置文件：cp conf/zoo_sample.cfg conf/zoo.cfg;

  vim conf/zoo.cfg

  dataDir=/usr/local/data/zookeeper-1

  clientPort=2181

  server.1=127.0.0.1:2888:3888

  server.2=127.0.0.1:2889:3889

  server.3=127.0.0.1:2890:3890

zoo.cfg复制两个配置文件（其只需修改dataDir和clientPort即可.） 

启动zk

创建集群根节点：/GroupMembers

  

服务提供与消费

public class ZkServiceConsumer implements Watcher {

    private static final Logger logger = LoggerFactory.getLogger(ZkServiceConsumer.class);

    private static CountDownLatch countDownLatch = new CountDownLatch(1);

    private static ZooKeeper zk = null;

    private static String root = "/GroupMembers";

    public static void main(String[] args) throws Exception {
        zk = new ZooKeeper("xxx.xxx.x.xx:2181", 5000, new ZkServiceConsumer());
        countDownLatch.await();
        List childrenList = zk.getChildren(root, true);
        logger.info("服务Provider:{}", childrenList);
        childrenList.forEach(child -> {
            try {
                logger.info(child + ":" + new String(zk.getData(root + "/" + child, false, new Stat())));
            } catch (KeeperException e) {
                e.printStackTrace();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });
        Thread.sleep(Integer.MAX_VALUE);
    }

    @Override
    public void process(WatchedEvent event) {
        if (Event.KeeperState.SyncConnected == event.getState()) {
            if (event.getType() == Event.EventType.None && event.getPath() == null) {
                countDownLatch.countDown();
            } else if (event.getType() == Event.EventType.NodeChildrenChanged) {
                try {
                    List childrenList = zk.getChildren(event.getPath(), true);
                    logger.info("更新服务提供者列表：{}", childrenList);
                    for (String child : childrenList) {
                        logger.info(child + new String(zk.getData(root + "/" + child, false, new Stat())));
                    }
                } catch (KeeperException e) {
                    e.printStackTrace();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            }
        }
    }
}

创建集群节点n1

  

创建集群节点n2

   同n1(port:8082)

public class ZkProSync implements Watcher {

    private static final Logger logger = LoggerFactory.getLogger(ZkProSync.class);

    private static CountDownLatch countDownLatch = new CountDownLatch(1);

    private static ZooKeeper zk = null;

    private static Stat stat = new Stat();

    public static void main(String[] args) throws Exception {
        String path = "/username";
        zk = new ZooKeeper("xxx.xxx.x.xx:2181", 5000, new ZkProSync());
        countDownLatch.await();
        logger.info(new String(zk.getData(path, true, stat)));
        Thread.sleep(Integer.MAX_VALUE);
    }

    @Override
    public void process(WatchedEvent event) {
        if (Event.KeeperState.SyncConnected == event.getState()) {
            if (event.getType() == Event.EventType.None && event.getPath() == null) {
                countDownLatch.countDown();
            } else if (event.getType() == Event.EventType.NodeDataChanged) {
                try {
                    logger.info("配置已更改为：{}", new String(zk.getData(event.getPath(), true, stat)));
                } catch (KeeperException e) {
                    e.printStackTrace();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            }
        }
    }
}

<其中Consistency：zk保证了最终一致性，Sync到各个节点.>

2、Eureka

  eureka不会有类似于zk选举leader的过程，若某台服务器宕机，客户端请求自动切换到新的eureka节点，当宕机的服务器恢复后，eureka则再次将其纳入到服务器集群管理之中，即同步新的服务注册信息。

1.CP VS 2.AP

ZK定位于分布式协调服务，在其管辖下的所有服务之间保持同步、一致(Zab算法，CP)，若作Service发现服务，其本身没有正确处理网络分割的问题<当多个zk之间网络出现问题-造成出现多个leader-脑裂>，即在同一个网络分区的节点数达不到zk选取leader的数目，它们就会从zk中断开，同时也不能提供Service发现服务l。

Eureka相对于ZK剔除了选取Leader或事务日志机制，它有独立的客户端程序库，同时提供心跳、服务健康监测、自动发布等服务与自动刷新缓存的功能，在网络分割故障发生时，每个节点会持续的对外提供服务，接收新的服务注册同时将它们提供给下游的服务发现请求(AP)l。