HDFS HA和Federation

时间 2019-11-07

标签 hdfs federation 栏目 Hadoop 繁體版

原文原文链接

cluster1和cluster2组成hdfs联盟，cluster1有namenodeA和namenodeB，cluster2有namenodeC和namenodeD。node

安装配置好后启动顺序应该为：bootstrap

1.在全部zookeeper集群节点上启动zookeeper：zkServer.sh start oop

2.在cluster1的namenodeA上格式化zookeeper集群：hdfs zkfc -formatZKspa

3.在cluster2的namenodeC上格式化zookeeper集群：hdfs zkfc -formatZK.net

4.在全部journal集群节点上启动journal：hadoop-daemon.sh start journalnodeorm

5.在cluster1的namenodeA上格式化集群：hdfs namenode -format -clusterId cluster1blog

6.启动cluster1的namenodeA：hadoop-daemon.sh start namenode教程

7.在cluster1的namenodeB上copynamenodeA的数据的数据：hdfs namenode -bootstrapStandbyhadoop

8.启动cluster1的namenodeB：hadoop-daemon.sh start namenodeget

9.在cluster2的namenodeC上格式化集群：hdfs namenode -format -clusterId cluster2

10.启动cluster2的namenodeC：hadoop-daemon.sh start namenode

11.在cluster2的namenodeD上copynamenodeC的数据的数据：hdfs namenode -bootstrapStandby

12.启动cluster2的namenodeD：hadoop-daemon.sh start namenode

13.启动全部namenode的zkfc：hadoop-daemon.sh start zkfc

14.启动全部的datanode：hadoop-daemon.sh start datanode

15.在resourcemanager上启动yarn：start-yarn.sh

16.验证HA，杀死active的namenode：kill -9

17.提交做业验证yarn验证。