集群搭建
因為服務器數量有限，這里服務器開啟的進程有點多：
| 機器名 | 安裝軟件 | 運行進程 |
| hadoop001 | Hadoop,Zookeeper | NameNode, DFSZKFailoverController, ResourceManager DataNode, NodeManager QuorumPeerMain JournalNode |
| hadoop002 | Hadoop,Zookeeper | NameNode, DFSZKFailoverController, ResourceManager DataNode, NodeManager QuorumPeerMain JournalNode |
| hadoop003 | Hadoop,Zookeeper | DataNode, NodeManager QuorumPeerMain |
說明[2]：
在hadoop2.X中通常由兩個NameNode組成，一個處於active狀態，另一個處於standby狀態。Active NameNode對外提供服務，而Standby NameNode則不對外提供服務，僅同步active namenode的狀態，以便能夠在它失敗時快速進行切換。
hadoop2.0官方提供了兩種HDFS HA的解決方案，一種是NFS，另一種是QJM（由cloudra提出，原理類似zookeeper）。這里我使用QJM完成。主備NameNode之間通過一組JournalNode同步元數據信息，一條數據只要成功寫入多數JournalNode即認為寫入成功。通常配置奇數個JournalNode
這里略去jdk，Hadoop，Zookeeper的安裝過程和環境變量配置。
無密碼登陸
這里要非常注意無密碼登陸的配置：
ssh-keygen -t rsa
在~/.ssh/目錄中生成兩個文件id_rsa和id_rsa.pub
如果想從hadoop001免密碼登錄到hadoop002中要在hadoop001中執行
ssh-copy-id -i ~/.ssh/id_rsa.pub [用戶名]@hadoop002
這里為了實現任何機器之間都可以免密碼登陸，所以在hadoop001中再執行兩遍上面的操作（把@后面的機器名分別改成hadoop001和hadoop003），最后把生成的authorized_keys復制所有的節點上
Hadoop配置
core-site.xml
<configuration> <!-- --> <property> <name>fs.defaultFS</name> <value>hdfs://appcluster</value> </property> <!-- 指定hadoop臨時目錄 --> <property> <name>hadoop.tmp.dir</name> <value>/data/hadoop/storage/tmp</value> </property> <!-- 指定zookeeper地址 --> <property> <name>ha.zookeeper.quorum</name> <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value> </property> <property> <name>ha.zookeeper.session-timeout.ms</name> <value>2000</value> </property> </configuration>
hdfs-site.xml
<configuration> <!--指定namenode名稱空間的存儲地址--> <property> <name>dfs.namenode.name.dir</name> <value>file:///data/hadoop/storage/hdfs/name</value> </property> <!--指定datanode數據存儲地址--> <property> <name>dfs.datanode.data.dir</name> <value>file:///data/hadoop/storage/hdfs/data</value> </property> <!--指定數據冗余份數--> <property> <name>dfs.replication</name> <value>2</value> </property> <!--指定hdfs的nameservice為appcluster，需要和core-site.xml中的保持一致 --> <property> <name>dfs.nameservices</name> <value>appcluster</value> </property> <!-- appcluster下面有兩個NameNode，分別是nn1，nn2 --> <property> <name>dfs.ha.namenodes.appcluster</name> <value>nn1,nn2</value> </property> <!-- nn1的RPC通信地址 --> <property> <name>dfs.namenode.rpc-address.appcluster.nn1</name> <value>hadoop001:8020</value> </property> <!-- nn2的RPC通信地址 --> <property> <name>dfs.namenode.rpc-address.appcluster.nn2</name> <value>hadoop002:8020</value> </property> <!-- nn1的http通信地址 --> <property> <name>dfs.namenode.http-address.appcluster.nn1</name> <value>hadoop001:50070</value> </property> <!-- nn2的http通信地址 --> <property> <name>dfs.namenode.http-address.appcluster.nn2</name> <value>hadoop002:50070</value> </property> <!-- 指定NameNode的元數據在JournalNode上的存放位置 --> <property> <name>dfs.namenode.shared.edits.dir</name> <value>qjournal://hadoop001:8485;hadoop002:8485;hadoop003:8485/appcluster</value> </property> <property> <name>dfs.ha.automatic-failover.enabled.appcluster</name> <value>true</value> </property> <!-- 配置失敗自動切換實現方式 --> <property> <name>dfs.client.failover.proxy.provider.appcluster</name> <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value> </property> <!-- 配置隔離機制 --> <property> <name>dfs.ha.fencing.methods</name> <value>sshfence</value> </property> <!-- 使用隔離機制時需要ssh免密碼登陸 --> <property> <name>dfs.ha.fencing.ssh.private-key-files</name> <value>/home/[用戶名]/.ssh/id_rsa</value> </property> <!-- --> <property> <name>dfs.journalnode.edits.dir</name> <value>/data/hadoop/tmp/journal</value> </property> </configuration>
mapred-site.xml
<configuration> <property> <name>mapreduce.framework.name</name> <value>yarn</value> </property> <!-- 配置 MapReduce JobHistory Server 地址 ，默認端口10020 --> <property> <name>mapreduce.jobhistory.address</name> <value>0.0.0.0:10020</value> </property> <!-- 配置 MapReduce JobHistory Server web ui 地址， 默認端口19888 --> <property> <name>mapreduce.jobhistory.webapp.address</name> <value>0.0.0.0:19888</value> </property> </configuration>
yarn-site.xml
<?xml version="1.0"?> <configuration> <!--rm失聯后重新鏈接的時間--> <property> <name>yarn.resourcemanager.connect.retry-interval.ms</name> <value>2000</value> </property> <!--開啟resourcemanagerHA,默認為false--> <property> <name>yarn.resourcemanager.ha.enabled</name> <value>true</value> </property> <!--配置resourcemanager--> <property> <name>yarn.resourcemanager.ha.rm-ids</name> <value>rm1,rm2</value> </property> <property> <name>ha.zookeeper.quorum</name> <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value> </property> <!--開啟故障自動切換--> <property> <name>yarn.resourcemanager.ha.automatic-failover.enabled</name> <value>true</value> </property> <property> <name>yarn.resourcemanager.hostname.rm1</name> <value>hadoop001</value> </property> <property> <name>yarn.resourcemanager.hostname.rm2</name> <value>hadoop002</value> </property> <!-- 在hadoop001上配置rm1,在hadoop002上配置rm2, 注意：一般都喜歡把配置好的文件遠程復制到其它機器上，但這個在YARN的另一個機器上一定要修改 --> <property> <name>yarn.resourcemanager.ha.id</name> <value>rm1</value> <description>If we want to launch more than one RM in single node,we need this configuration</description> </property> <!--開啟自動恢復功能--> <property> <name>yarn.resourcemanager.recovery.enabled</name> <value>true</value> </property> <!--配置與zookeeper的連接地址--> <property> <name>yarn.resourcemanager.zk-state-store.address</name> <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value> </property> <property> <name>yarn.resourcemanager.store.class</name> <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value> </property> <property> <name>yarn.resourcemanager.zk-address</name> <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value> </property> <property> <name>yarn.resourcemanager.cluster-id</name> <value>appcluster-yarn</value> </property> <!--schelduler失聯等待連接時間--> <property> <name>yarn.app.mapreduce.am.scheduler.connection.wait.interval-ms</name> <value>5000</value> </property> <!--配置rm1--> <property> <name>yarn.resourcemanager.address.rm1</name> <value>hadoop001:8032</value> </property> <property> <name>yarn.resourcemanager.scheduler.address.rm1</name> <value>hadoop001:8030</value> </property> <property> <name>yarn.resourcemanager.webapp.address.rm1</name> <value>hadoop001:8088</value> </property> <property> <name>yarn.resourcemanager.resource-tracker.address.rm1</name> <value>hadoop001:8031</value> </property> <property> <name>yarn.resourcemanager.admin.address.rm1</name> <value>hadoop001:8033</value> </property> <property> <name>yarn.resourcemanager.ha.admin.address.rm1</name> <value>hadoop001:23142</value> </property> <!--配置rm2--> <property> <name>yarn.resourcemanager.address.rm2</name> <value>hadoop002:8032</value> </property> <property> <name>yarn.resourcemanager.scheduler.address.rm2</name> <value>hadoop002:8030</value> </property> <property> <name>yarn.resourcemanager.webapp.address.rm2</name> <value>hadoop002:8088</value> </property> <property> <name>yarn.resourcemanager.resource-tracker.address.rm2</name> <value>hadoop002:8031</value> </property> <property> <name>yarn.resourcemanager.admin.address.rm2</name> <value>hadoop002:8033</value> </property> <property> <name>yarn.resourcemanager.ha.admin.address.rm2</name> <value>hadoop002:23142</value> </property> <property> <name>yarn.nodemanager.aux-services</name> <value>mapreduce_shuffle</value> </property> <property> <name>yarn.nodemanager.aux-services.mapreduce.shuffle.class</name> <value>org.apache.hadoop.mapred.ShuffleHandler</value> </property> <property> <name>yarn.nodemanager.local-dirs</name> <value>/data/hadoop/yarn/local</value> </property> <property> <name>yarn.nodemanager.log-dirs</name> <value>/data/hadoop/yarn/log</value> </property> <property> <name>mapreduce.shuffle.port</name> <value>23080</value> </property> <!--故障處理類--> <property> <name>yarn.client.failover-proxy-provider</name> <value>org.apache.hadoop.yarn.client.ConfiguredRMFailoverProxyProvider</value> </property> <property> <name>yarn.resourcemanager.ha.automatic-failover.zk-base-path</name> <value>/yarn-leader-election</value> <description>Optionalsetting.Thedefaultvalueis/yarn-leader-election</description> </property> </configuration>
hadoop-env.sh & mapred-env.sh & yarn-env.sh
export JAVA_HOME=/usr/java/jdk1.7.0_60
export CLASS_PATH=$JAVA_HOME/lib:$JAVA_HOME/jre/lib
export HADOOP_HOME=/data/hadoop-2.6.0
export HADOOP_PID_DIR=/data/hadoop/pids
export HADOOP_COMMON_LIB_NATIVE_DIR=$HADOOP_HOME/lib/native
export HADOOP_OPTS="$HADOOP_OPTS-Djava.library.path=$HADOOP_HOME/lib/native"
export HADOOP_PREFIX=$HADOOP_HOME
export HADOOP_MAPRED_HOME=$HADOOP_HOME
export HADOOP_COMMON_HOME=$HADOOP_HOME
export HADOOP_HDFS_HOME=$HADOOP_HOME
export YARN_HOME=$HADOOP_HOME
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export HDFS_CONF_DIR=$HADOOP_HOME/etc/hadoop
export YARN_CONF_DIR=$HADOOP_HOME/etc/hadoop
export JAVA_LIBRARY_PATH=$HADOOP_HOME/lib/native
export PATH=$PATH:$JAVA_HOME/bin:$HADOOP_HOME/bin:$HADOOP_HOME/sbin
參考文獻
[1] hdfs-site.xml:http://www.21ops.com/front-tech/10744.html
[2] yarn-site.xml: http://www.aboutyun.com/thread-10572-1-1.html 評論也值得參考
僅參考這兩篇配置后報錯：
15/07/17 13:58:55 FATAL ha.ZKFailoverController: Automatic failover is not enabled for NameNode at hadoop001/**.**.**.**:8020. Please ensure that automatic failover is enabled in the configuration before running the ZK failover controller.
再參考
[3]http://www.cnblogs.com/meiyuanbao/p/3545929.html (沒有做到Yarn的HA)
發現需要在hdfs-site.xml添加配置：
<property> <name>dfs.ha.automatic-failover.enabled.appcluster</name> <value>true</value> </property>