2.11核心文件配置

　　注：这里特别提醒，配置文件中的路径在启动集群之前，得存在（若不存在，请事先创建）。下面为给出本篇文章需要创建的路径脚本，命令如下：

1. mkdir -p /home/hadoop/tmp
2. mkdir -p /home/hadoop/data/tmp/journal
3. mkdir -p /home/hadoop/data/dfs/name
4. mkdir -p /home/hadoop/data/dfs/data
5. mkdir -p /home/hadoop/data/yarn/local
6. mkdir -p /home/hadoop/log/yarn

• core-site.xml

1. <?xml version="1.0" encoding="UTF-8"?>
2. <configuration>
3. <property>
4. <name>fs.defaultFS</name>
5. <value>hdfs://cluster1</value>
6. </property>
7.  
8. <property>
9. <name>io.file.buffer.size</name>
10. <value>131072</value>
11. </property>
12. <property>
13. <name>hadoop.tmp.dir</name>
14. <value>/home/hadoop/tmp</value>
15. </property>
16. <property>
17. <name>hadoop.proxyuser.hduser.hosts</name>
18. <value>*</value>
19. </property>
20. <property>
21. <name>hadoop.proxyuser.hduser.groups</name>
22. <value>*</value>
23. </property>
24. <property>
25. <name>ha.zookeeper.quorum</name>
26. <value>dn1:2181,dn2:2181,dn3:2181</value>
27. </property>
28. </configuration>

• hdfs-site.xml

1. <?xml version="1.0" encoding="UTF-8"?>
2. <configuration>
3. <property>
4. <name>dfs.nameservices</name>
5. <value>cluster1</value>
6. </property>
7. <property>
8. <name>dfs.ha.namenodes.cluster1</name>
9. <value>nna,nns</value>
10. </property>
11. <property>
12. <name>dfs.namenode.rpc-address.cluster1.nna</name>
13. <value>nna:9000</value>
14. </property>
15. <property>
16. <name>dfs.namenode.rpc-address.cluster1.nns</name>
17. <value>nns:9000</value>
18. </property>
19.  
20. <property>
21. <name>dfs.namenode.http-address.cluster1.nna</name>
22. <value>nna:50070</value>
23. </property>
24.  
25. <property>
26. <name>dfs.namenode.http-address.cluster1.nns</name>
27. <value>nns:50070</value>
28. </property>
29. <property>
30. <name>dfs.namenode.shared.edits.dir</name>
31. <value>qjournal://dn1:8485;dn2:8485;dn3:8485/cluster1</value>
32. </property>
33.  
34. <property>
35. <name>dfs.client.failover.proxy.provider.cluster1</name>
36. <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
37. </property>
38. <property>
39. <name>dfs.ha.fencing.methods</name>
40. <value>sshfence</value>
41. </property>
42. <property>
43. <name>dfs.ha.fencing.ssh.private-key-files</name>
44. <value>/home/hadoop/.ssh/id_rsa</value>
45. </property>
46. <property>
47. <name>dfs.journalnode.edits.dir</name>
48. <value>/home/hadoop/data/tmp/journal</value>
49. </property>
50. <property>
51. <name>dfs.ha.automatic-failover.enabled</name>
52. <value>true</value>
53. </property>
54. <property>
55. <name>dfs.namenode.name.dir</name>
56. <value>/home/hadoop/data/dfs/name</value>
57. </property>
58. <property>
59. <name>dfs.datanode.data.dir</name>
60. <value>/home/hadoop/data/dfs/data</value>
61. </property>
62. <property>
63. <name>dfs.replication</name>
64. <value>3</value>
65. </property>
66. <property>
67. <name>dfs.webhdfs.enabled</name>
68. <value>true</value>
69. </property>
70.  
71. <property>
72. <name>dfs.journalnode.http-address</name>
73. <value>0.0.0.0:8480</value>
74. </property>
75. <property>
76. <name>dfs.journalnode.rpc-address</name>
77. <value>0.0.0.0:8485</value>
78. </property>
79. <property>
80. <name>ha.zookeeper.quorum</name>
81. <value>dn1:2181,dn2:2181,dn3:2181</value>
82. </property>
83.  
84. </configuration>

• map-site.xml

1. <?xml version="1.0" encoding="UTF-8"?>
2. <configuration>
3. <property>
4. <name>mapreduce.framework.name</name>
5. <value>yarn</value>
6. </property>
7. <property>
8. <name>mapreduce.jobhistory.address</name>
9. <value>nna:10020</value>
10. </property>
11. <property>
12. <name>mapreduce.jobhistory.webapp.address</name>
13. <value>nna:19888</value>
14. </property>
15. </configuration>

• yarn-site.xml

1. <?xml version="1.0" encoding="UTF-8"?>
2. <configuration>
3. <property>
4. <name>yarn.resourcemanager.connect.retry-interval.ms</name>
5. <value>2000</value>
6. </property>
7. <property>
8. <name>yarn.resourcemanager.ha.enabled</name>
9. <value>true</value>
10. </property>
11. <property>
12. <name>yarn.resourcemanager.ha.rm-ids</name>
13. <value>rm1,rm2</value>
14. </property>
15. <property>
16. <name>ha.zookeeper.quorum</name>
17. <value>dn1:2181,dn2:2181,dn3:2181</value>
18. </property>
19.  
20. <property>
21. <name>yarn.resourcemanager.ha.automatic-failover.enabled</name>
22. <value>true</value>
23. </property>
24. <property>
25. <name>yarn.resourcemanager.hostname.rm1</name>
26. <value>nna</value>
27. </property>
28.  
29. <property>
30. <name>yarn.resourcemanager.hostname.rm2</name>
31. <value>nns</value>
32. </property>
33. <!--在namenode1上配置rm1,在namenode2上配置rm2,注意：一般都喜欢把配置好的文件远程复制到其它机器上，但这个在YARN的另一个机器上一定要修改 -->
34. <property>
35. <name>yarn.resourcemanager.ha.id</name>
36. <value>rm1</value>
37. </property>
38. <!--开启自动恢复功能 -->
39. <property>
40. <name>yarn.resourcemanager.recovery.enabled</name>
41. <value>true</value>
42. </property>
43. <!--配置与zookeeper的连接地址 -->
44. <property>
45. <name>yarn.resourcemanager.zk-state-store.address</name>
46. <value>dn1:2181,dn2:2181,dn3:2181</value>
47. </property>
48. <property>
49. <name>yarn.resourcemanager.store.class</name>
50. <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value>
51. </property>
52. <property>
53. <name>yarn.resourcemanager.zk-address</name>
54. <value>dn1:2181,dn2:2181,dn3:2181</value>
55. </property>
56. <property>
57. <name>yarn.resourcemanager.cluster-id</name>
58. <value>cluster1-yarn</value>
59. </property>
60. <!--schelduler失联等待连接时间 -->
61. <property>
62. <name>yarn.app.mapreduce.am.scheduler.connection.wait.interval-ms</name>
63. <value>5000</value>
64. </property>
65. <!--配置rm1 -->
66. <property>
67. <name>yarn.resourcemanager.address.rm1</name>
68. <value>nna:8132</value>
69. </property>
70. <property>
71. <name>yarn.resourcemanager.scheduler.address.rm1</name>
72. <value>nna:8130</value>
73. </property>
74. <property>
75. <name>yarn.resourcemanager.webapp.address.rm1</name>
76. <value>nna:8188</value>
77. </property>
78. <property>
79. <name>yarn.resourcemanager.resource-tracker.address.rm1</name>
80. <value>nna:8131</value>
81. </property>
82. <property>
83. <name>yarn.resourcemanager.admin.address.rm1</name>
84. <value>nna:8033</value>
85. </property>
86. <property>
87. <name>yarn.resourcemanager.ha.admin.address.rm1</name>
88. <value>nna:23142</value>
89. </property>
90. <!--配置rm2 -->
91. <property>
92. <name>yarn.resourcemanager.address.rm2</name>
93. <value>nns:8132</value>
94. </property>
95. <property>
96. <name>yarn.resourcemanager.scheduler.address.rm2</name>
97. <value>nns:8130</value>
98. </property>
99. <property>
100. <name>yarn.resourcemanager.webapp.address.rm2</name>
101. <value>nns:8188</value>
102. </property>
103. <property>
104. <name>yarn.resourcemanager.resource-tracker.address.rm2</name>
105. <value>nns:8131</value>
106. </property>
107. <property>
108. <name>yarn.resourcemanager.admin.address.rm2</name>
109. <value>nns:8033</value>
110. </property>
111. <property>
112. <name>yarn.resourcemanager.ha.admin.address.rm2</name>
113. <value>nns:23142</value>
114. </property>
115. <property>
116. <name>yarn.nodemanager.aux-services</name>
117. <value>mapreduce_shuffle</value>
118. </property>
119. <property>
120. <name>yarn.nodemanager.aux-services.mapreduce.shuffle.class</name>
121. <value>org.apache.hadoop.mapred.ShuffleHandler</value>
122. </property>
123. <property>
124. <name>yarn.nodemanager.local-dirs</name>
125. <value>/home/hadoop/data/yarn/local</value>
126. </property>
127. <property>
128. <name>yarn.nodemanager.log-dirs</name>
129. <value>/home/hadoop/log/yarn</value>
130. </property>
131. <property>
132. <name>mapreduce.shuffle.port</name>
133. <value>23080</value>
134. </property>
135. <!--故障处理类 -->
136. <property>
137. <name>yarn.client.failover-proxy-provider</name>
138. <value>org.apache.hadoop.yarn.client.ConfiguredRMFailoverProxyProvider</value>
139. </property>
140. <property>
141. <name>yarn.resourcemanager.ha.automatic-failover.zk-base-path</name>
142. <value>/yarn-leader-election</value>
143. </property>
144. </configuration>

• hadoop-env.sh

1. # The java implementation to use.
2. export JAVA_HOME=/usr/java/jdk1.7

• yarn-env.sh

1. # some Java parameters
2. export JAVA_HOME=/usr/java/jdk1.7

• 2.12slave

　　修改hadoop安装目录下的slave文件：

1. dn1
2. dn2
3. dn3

2.13启动命令（hdfs和yarn的相关命令）

　　由于我们配置了QJM，所以我们需要先启动QJM的服务，启动顺序如下所示：

1. 进入到DN节点，启动zk的服务：zkServer.sh start，之后可以输入zkServer.sh status查看启动状态，本次我们配置了三个DN节点，会出现一个leader和两个follower。输入jps，会显示启动进程：QuorumPeerMain
2. 在NN节点上（选一台即可，这里我选择的是一台预NNA节点），然后启动journalnode服务，命令如下：hadoop-daemons.sh start journalnode。或者单独进入到每个DN输入启动命令：hadoop-daemon.sh start journalnode。输入jps显示启动进程：JournalNode。
3. 接着若是配置后，我们首次启动，需要格式化HDFS，命令如下：hadoop namenode –format。
4. 之后我们需要格式化ZK，命令如下：hdfs zkfc –formatZK。
5. 接着我们启动hdfs和yarn，命令如下：start-dfs.sh和start-yarn.sh，我们在nna输入jps查看进程，显示如下：DFSZKFailoverController，NameNode，ResourceManager。
6. 接着我们在NNS输入jps查看，发现只有DFSZKFailoverController进程，这里我们需要手动启动NNS上的namenode和ResourceManager进程，命令如下：hadoop-daemon.sh start namenode和yarn-daemon.sh start resourcemanager。需要注意的是，在NNS上的yarn-site.xml中，需要配置指向NNS，属性配置为rm2，在NNA中配置的是rm1。
7. 最后我们需要同步NNA节点的元数据，命令如下：hdfs namenode –bootstrapStandby，若执行正常，日志最后显示如下信息：

1. 15/02/21 10:30:59 INFO common.Storage: Storage directory /home/hadoop/data/dfs/name has been successfully formatted.
2. 15/02/21 10:30:59 WARN common.Util: Path /home/hadoop/data/dfs/name should be specified as a URI in configuration files. Please update hdfs configuration.
3. 15/02/21 10:30:59 WARN common.Util: Path /home/hadoop/data/dfs/name should be specified as a URI in configuration files. Please update hdfs configuration.
4. 15/02/21 10:31:00 INFO namenode.TransferFsImage: Opening connection to http://nna:50070/imagetransfer?getimage=1&txid=0&storageInfo=-60:1079068934:0:CID-1dd0c11e-b27e-4651-aad6-73bc7dd820bd
5. 15/02/21 10:31:01 INFO namenode.TransferFsImage: Image Transfer timeout configured to 60000 milliseconds
6. 15/02/21 10:31:01 INFO namenode.TransferFsImage: Transfer took 0.01s at 0.00 KB/s
7. 15/02/21 10:31:01 INFO namenode.TransferFsImage: Downloaded file fsimage.ckpt_0000000000000000000 size 353 bytes.
8. 15/02/21 10:31:01 INFO util.ExitUtil: Exiting with status 0
9. 15/02/21 10:31:01 INFO namenode.NameNode: SHUTDOWN_MSG: /************************************************************ SHUTDOWN_MSG: Shutting down NameNode at nns/10.211.55.13 ************************************************************/

2.14HA的切换

　　由于我配置的是自动切换，若NNA节点宕掉，NNS节点会立即由standby状态切换为active状态。若是配置的手动状态，可以输入如下命令进行人工切换：

1. hdfs haadmin -failover --forcefence --forceactive nna nns

　　这条命令的意思是，将nna变成standby，nns变成active。而且手动状态下需要重启服务。

2.15效果截图

3.总结

　　这篇文章就赘述到这里，若在配置过程中有什么疑问或问题，可以加入QQ群讨论或发送邮件给我，我会尽我所能为您解答，与君共勉！