架構

Hadoop叢集的啟動

zookeeper:
/home/pubsrv/zookeeper-3.4.3-cdh4.0.0/bin/zkServer.sh start   //1,2,3,4,5

jobhistoryserver:(自帶的一個歷史伺服器，可以通過歷史伺服器檢視已經執行完的Mapreduce作業記錄)
/home/pubsrv/hadoop-cdh5/lib/hadoop-mapreduce/sbin/mr-jobhistory-daemon.sh start historyserver   //1

resourcemanager:
先jps，並kill掉resourcemanager
/home/pubsrv/hadoop-cdh5/lib/hadoop-yarn/sbin/yarn-daemon.sh start resourcemanager    //1,2  stop(
 
停止)

namenode:
/home/pubsrv/hadoop-cdh5/lib/hadoop/sbin/hadoop-daemon.sh start namenode       //1,2

zkfc:(保證NameNode的高可用，當一個NameNode有異常時，斷開與zookeeper的連線)
/home/pubsrv/hadoop-cdh5/lib/hadoop/sbin/hadoop-daemon.sh start zkfc      //1,2

journalnode:（2個NameNode共享資料）
/home/pubsrv/hadoop-cdh5/lib/hadoop/sbin/hadoop-daemon.sh start journalnode     //1,2,3,4,5

nodemanger:
/home/pubsrv/hadoop-cdh5/lib/hadoop-yarn/sbin/yarn-daemon.sh start nodemanager     //2-28

datanode:
 /home/pubsrv/hadoop-cdh5/lib/hadoop/sbin/hadoop-daemon.sh start datanode    //2-28

!
 
[在這裡插入圖片描述](https://img-blog.csdnimg.cn/20181217120351290.png)

參考部落格

Hadoop叢集的JobHistoryServer詳解
[1]https://blog.csdn.net/play_chess_ITmanito/article/details/51965279
NameNode的ZKFC機制
[2]https://www.cnblogs.com/lixiaolun/p/6897706.html
Hadoop Journal Node 作用
[3]https://my.oschina.net/u/189445/blog/661561