[关闭]
@duguyiren3476 2015-08-04T08:37:40.000000Z 字数 4665 阅读 331

Hadoop HA 高可用配置

未分类

hadoop的ha高可用是指hdfs的高可用机制,如果要配置ha,请在配置好的非高可用的配置基础上修改,环境
- namenode:
nn1:10.128.17.21;
nn2:10.128.17.39;
- datanode:
dn1:10.128.17.24
dn2:10.128.17.25
- journalnode:
10.128.17.39;
10.128.17.24;
10.128.17.25;
- zookeeper:
10.128.17.16
10.128.17.17
10.128.17.20
如下文件:

core-site.xml

  1. <configuration>
  2.         <property>
  3.                 <name>hadoop.tmp.dir</name>
  4.                 <value>/oneapm/data/hadoop</value>
  5.                 <description>Abase for other temporary directories.</description>
  6.         </property>
  7.         <property>
  8.                 <name>fs.defaultFS</name>
  9.                 <value>hdfs://onlinecluster</value>
  10.         </property>
  11.         <property>
  12.                 <name>fs.trash.interval</name>
  13.                 <value>2880</value>
  14.                 <description>trash save two days (m)</description>
  15.         </property>
  16.         <property>
  17.                 <name>hadoop.logfile.size</name>
  18.                 <value>104857600</value>
  19.                 <description>The max size of each log file 100M</description>
  20.         </property>
  21.         <property>
  22.                 <name>hadoop.logfile.count</name>
  23.                 <value>10</value>
  24.                 <description>The max number of log files</description>
  25.         </property> 
  26.         <property>
  27.                 <name>hadoop.native.lib</name>
  28.                 <value>true</value>
  29.         </property> 
  30.         <property> 
  31.                 <name>io.compression.codecs</name>
  32.                 <value>
  33.                 org.apache.hadoop.io.compress.GzipCodec,
  34.                 org.apache.hadoop.io.compress.DefaultCodec,
  35.                 org.apache.hadoop.io.compress.BZip2Codec,
  36.                 org.apache.hadoop.io.compress.SnappyCodec
  37.                 </value>
  38.          </property>
  39.          <property> 
  40.                 <name>ha.zookeeper.quorum</name>
  41.                 <value>10.128.17.16:2181,10.128.17.17:2181,10.128.17.20:2181</value>
  42.         </property>
  43.         <property>
  44.                 <name>ha.zookeeper.session-timeout.ms</name>
  45.                 <value>1000</value>
  46.                 <description>ms</description>
  47.         </property>
  48. </configuration>

hdfs-site.xml

  1. <configuration>
  2.     <property>
  3.         <name>dfs.replication</name>
  4.         <value>2</value>
  5.     </property>
  6.     <property>
  7.         <name>dfs.support.append</name>
  8.         <value>true</value>
  9.     </property>
  10.     <property>
  11.         <name>dfs.permissions</name>
  12.         <value>false</value>
  13.     </property>
  14.     <property>
  15.         <name>dfs.nameservices</name>
  16.         <value>onlinecluster</value>
  17.     </property>
  18.     <property>
  19.         <name>dfs.ha.namenodes.onlinecluster</name>
  20.         <value>nn1,nn2</value>
  21.     </property>
  22.     <property>
  23.         <name>dfs.namenode.rpc-address.onlinecluster.nn1</name>
  24.         <value>10.128.17.21:8020</value>
  25.     </property>
  26.     <property>
  27.         <name>dfs.namenode.rpc-address.onlinecluster.nn2</name>
  28.         <value>10.128.17.39:8020</value>
  29.     </property>
  30.     <property>
  31.         <name>dfs.namenode.servicerpc-address.onlinecluster.nn1</name>
  32.         <value>10.128.17.21:53310</value>
  33.     </property>
  34.     <property>
  35.         <name>dfs.namenode.servicerpc-address.onlinecluster.nn2</name>
  36.         <value>10.128.17.39:53310</value>
  37.     </property>
  38.     <property>
  39.         <name>dfs.namenode.http-address.onlinecluster.nn1</name>
  40.         <value>10.128.17.21:50070</value>
  41.     </property>
  42.     <property>
  43.         <name>dfs.namenode.http-address.onlinecluster.nn2</name>
  44.         <value>10.128.17.39:50070</value>
  45.     </property>
  46.     <property>
  47.         <name>dfs.namenode.shared.edits.dir</name>
  48.         <value>qjournal://10.128.17.24:8485;10.128.17.25:8485;1 0.128.17.39:8485/onlinecluster</value>
  49.     </property>
  50.     <property>
  51.         <name>dfs.client.failover.proxy.provider.onlinecluster</name>
  52.         <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
  53.     </property>
  54.     <property>
  55.         <name>dfs.ha.fencing.methods</name>
  56.         <value>sshfence</value>
  57.     </property>
  58.     <property>
  59.         <name>dfs.ha.fencing.ssh.private-key-files</name>
  60.         <value>/home/hadoop/.ssh/id_rsa_nn1</value>
  61.     </property>
  62.     <property>
  63.         <name>dfs.ha.fencing.ssh.connect-timeout</name>
  64.         <value>30000</value>
  65.     </property>
  66.     <property>
  67.         <name>dfs.journalnode.edits.dir</name>
  68.         <value>/oneapm/data/hadoop/journaldata</value>
  69.     </property>
  70.     <property>
  71.         <name>dfs.ha.automatic-failover.enabled</name>
  72.         <value>true</value>
  73.     </property>
  74.     <property>
  75.         <name>ha.failover-controller.cli-check.rpc-timeout.ms</name>
  76.         <value>60000</value>
  77.     </property>
  78.     <property>
  79.         <name>ipc.client.connect.timeout</name>
  80.         <value>60000</value>
  81.     </property>
  82.     <property>
  83.         <name>dfs.image.transfer.bandwidthPerSec</name>
  84.         <value>4194304</value>
  85.     </property>
  86. </configuration>

--
dfs.ha.fencing.methods 的方式为ssh方式,也可以选择shell方式,后者相比更加的麻烦
dfs.ha.fencing.ssh.private-key-files :/home/hadoop/.ssh/id_rsa_nn1 采用sshfence方式的话,需要nn1和nn2之间相互copy对方的私钥id_rsa 并重命名为nn1和nn2; 配置文件同步完之后此值要相应修改为id_rsa_nn1和id_rsa_nn2
同时需要在nn1和nn2的主机上吧所有的dn哈nn的hosts文件配置完整:之后scp分发配置到各个节点

安装启动

  1. bin/hdfs namenode -bootstrapStandby
  1. sbin/stop-dfs.sh
  1. sbin/start-dfs.sh

浏览器粉笔访问nn1:50070和nn2:50070查看namenode状态,并分别kill掉active状态的namenode进程,观察是否standby状态的namenode改变为active,如此反复测试ha

添加新批注
在作者公开此批注前，只有你和作者可见。
回复批注