zookeeper选举问题

两种选举制度

1.持久-直到节点挂掉

2.轮留选举

这里我使用的是持久的方式

public class ZKConnectionUtil {

    private  static CuratorFramework client ;

    private static Object lock = new Object();

    private static LeaderLatch leaderLatch;

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

    /**

     * @描述：是否是leader

     * @return

     * @return boolean

     * @exception

     * @createTime：2016年11月22日

     * @author: songqinghu

     */

    public static boolean isLeader(CuratorFramework client){

        return leaderLatch.hasLeadership();

    }

//            solr.zk.check.namespace=contentcheck

//            solr.zk.check.flag.path=/flag

    /**

     * 

     * @描述：获取单例的zookeeper连接

     * @return void

     * @exception

     * @createTime：2016年8月24日

     * @author: songqinghu

     */

    public static CuratorFramework getZKConnection(){

        if(client ==null){

            synchronized (lock) {

                if(client ==null){

                    int connectionTimeoutMs=5000;

                    String connectString=PropertiesUtil.getProProperties("solr.zk.url","");

                    String namespace=PropertiesUtil.getProProperties("solr.zk.porsche.namespace","");

                    //String connectString="10.125.2.48:3181";

                    //String namespace="contentcheck";

                    //byte[] defaultData=IpUtil.getInNetworkIp().getBytes();

                    RetryPolicy retryPolicy= new RetryNTimes(Integer.MAX_VALUE, 1000);

                    client = CuratorFrameworkFactory.

                            builder().

//        aclProvider(aclProvider).

//        authorization(scheme, auth).

                            connectionTimeoutMs(connectionTimeoutMs).

                            connectString(connectString).

                          //  defaultData(defaultData).

                            namespace(namespace).

                            retryPolicy(retryPolicy).

                            build();

                    client.start();

                    //选举

                    String leaderPath = PropertiesUtil.getProProperties("solr.zk.porsche.leader", "");

                    leaderLatch = new LeaderLatch(client, leaderPath);

                    try {

                        leaderLatch.start();

                    } catch (Exception e) {

                        logger.error("get leaderlatch occor error and error is : " + e);    

                    }  

                }

            }

        }

        return client;

    }

}

原理解析:
在指定节点下每台机器会创建临时节点,机器挂掉,节点会被删除,就会重新选举