cluster1和cluster2組成hdfs聯盟,cluster1有namenodeA和namenodeB,cluster2有namenodeC和namenodeD。node
安裝配置好後啓動順序應該爲:bootstrap
1.在全部zookeeper集羣節點上啓動zookeeper:zkServer.sh start oop
2.在cluster1的namenodeA上格式化zookeeper集羣:hdfs zkfc -formatZKspa
3.在cluster2的namenodeC上格式化zookeeper集羣:hdfs zkfc -formatZK.net
4.在全部journal集羣節點上啓動journal:hadoop-daemon.sh start journalnodeorm
5.在cluster1的namenodeA上格式化集羣:hdfs namenode -format -clusterId cluster1blog
6.啓動cluster1的namenodeA:hadoop-daemon.sh start namenode教程
7.在cluster1的namenodeB上copynamenodeA的數據的數據:hdfs namenode -bootstrapStandbyhadoop
8.啓動cluster1的namenodeB:hadoop-daemon.sh start namenodeget
9.在cluster2的namenodeC上格式化集羣:hdfs namenode -format -clusterId cluster2
10.啓動cluster2的namenodeC:hadoop-daemon.sh start namenode
11.在cluster2的namenodeD上copynamenodeC的數據的數據:hdfs namenode -bootstrapStandby
12.啓動cluster2的namenodeD:hadoop-daemon.sh start namenode
13.啓動全部namenode的zkfc:hadoop-daemon.sh start zkfc
14.啓動全部的datanode:hadoop-daemon.sh start datanode
15.在resourcemanager上啓動yarn:start-yarn.sh
16.驗證HA,殺死active的namenode:kill -9
17.提交做業驗證yarn驗證。