Hadoop2分布式及NN和RM实现HA的实验

时间:2016-06-10 01:10 来源:linux.it.net.cn 作者:it.net.cn

目录结构

引言
实验环境
实验过程
演示demo

[一]、引言

在Hadoop2.x初期的时候写过一篇 hadoop 2.2.0 集群模式安装配置和测试，记录了分布式搭建的最基本的搭建步骤和运行演示，那篇文章中没有对HA的配置做实验，本文会详细介绍 Hadoop2的分布式、NameNode配置HA以及ResourceManage配置HA的实验过程。

[二]、实验环境

1、各节点及角色分配

本文以5个集群节点为基础做实验环境，具体的角色分配如下：

hostname	NameNode	DataNode	JournalNode	Zookeeper	ZKFC	ResourceManager
nn1.hadoop	√(Active)		√	√	√	√
nn2.hadoop	√(Standby)		√	√	√	√
dn1.hadoop		√	√	√
dn2.hadoop		√
dn3.hadoop		√

2、系统及软件版本

CentOS 6.3 64位
Java 1.7.0_75
Hadoop 2.6.0
zookeeper 3.4.6

3、安装JDK （所有节点需要操作）

//查询openjdk相关安装包

rpm -qa | grep java

//卸载openjdk

rpm -e --nodeps java-1.6.0-openjdk-1.6.0.0-1.45.1.11.1.el6.x86_64

rpm -e --nodeps java-1.6.0-openjdk-javadoc-1.6.0.0-1.45.1.11.1.el6.x86_64

rpm -e --nodeps java-1.6.0-openjdk-devel-1.6.0.0-1.45.1.11.1.el6.x86_64

rpm -e --nodeps tzdata-java-2012c-1.el6.noarch

rpm -e --nodeps java-1.5.0-gcj-1.5.0.0-29.1.el6.x86_64

Oracle官方下载 64为 jdk ：jdk-7u3-linux-x64.rpm 执行安装命令：

rpm -ivh jdk-7u3-linux-x64.rpm

默认的安装路径：/usr/java/jdk1.7.0_75

4、配置hosts （所有节点需要操作）

172.17.225.61 nn1.hadoop zk1.hadoop

172.17.225.121 nn2.hadoop zk2.hadoop

172.17.225.72 dn1.hadoop zk3.hadoop

172.17.225.76 dn2.hadoop

172.17.225.19 dn3.hadoop

5、确认SSHD已经安装并启动（所有节点需要操作）

6、配置时钟同步

第一种方法：（所有节点都要操作）都从公共NTP服务器同步，执行如下：

$ cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime

$ ntpdate us.pool.ntp.org

$ crontab -e

0-59/10 * * * * /usr/sbin/ntpdate us.pool.ntp.org | logger -t NTP

第二种方法：选一个节点搭建一个NTP服务，其他节点从该NTP服务器同步

7、创建专有用户（所有节点需要操作）

比如创建 hadoop用户，密码也初始化为hadoop, 下面有关hadoop部署配置都是以这个用户操作的

groupadd hadoop

useradd -g hadoop hadoop

passwd hadoop

为hadoop 用户修改环境变量 vi ~/.bash_profile ：

1 2	export JAVA_HOME=/usr/java/jdk1.7.0_75 export PATH="$JAVA_HOME/bin:$PATH"

8、SSH免密码登陆

配置所有的NameNode节点可以免密码登录到其余所有节点，只需要单向免密登录即可，当然你要配置为双向也无妨。有关SSH无密码登录的详细介绍可以参见：Linux(Centos)配置OpenSSH无密码登陆

[三]、实验过程

1、hadoop2的编译

在实验环境中任节点机器上下载hadoop 2.6.0的源码，安装配置好Java 和Maven 然后执行 mvn package -Pdist,native -DskipTests -Dtar 进行源码编译，具体可参考：

Hadoop2.2.0源码编译
Hadoop2.x在Ubuntu系统中编译源码

2、zookeeper安装配置

下载最新稳定版本（3.4.6）部署在ZK的各个节点，修改环境变量vi ~/.bash_profile ：

1 2	export ZOOKEEPER_HOME=/usr/local/share/zookeeper export PATH="$ZOOKEEPER_HOME/bin:$PATH"

修改配置文件：

cd $ZOOKEEPER_HOME

cp conf/zoo_sample.cfg conf/zoo.cfg

vi conf/zoo.cfg

修改成如下：

tickTime=2000

initLimit=10

syncLimit=5

clientPort=2181

dataDir=/bigdata/hadoop/zookeeper/zkdata

dataLogDir=/bigdata/hadoop/zookeeper/zklogs

server.1=zk1.hadoop:2888:3888

server.2=zk2.hadoop:2888:3888

server.3=zk3.hadoop:2888:3888

配置文件中的相关目录路径需要先创建好且hadoop用户具有读写权限,不同zk节点配置不同的myid：

在zk1.hadoop 节点中执行：echo 1 > /bigdata/hadoop/zookeeper/zkdata/myid
在zk2.hadoop 节点中执行：echo 2 > /bigdata/hadoop/zookeeper/zkdata/myid
在zk3.hadoop 节点中执行：echo 3 > /bigdata/hadoop/zookeeper/zkdata/myid

myid中的数值需要和 zoo.cfg中的配置一致。

3、hadoop 安装配置（所有节点需要修改）

3.1、配置环境变量vi ~/.bash_profile ：

1 2	export HADOOP_HOME=/usr/local/share/hadoop export PATH="$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin"

3.2、修改 $HADOOP_HOME/etc/hadoop/core-site.xml

<name>fs.defaultFS</name>

<value>hdfs://mycluster</value>

<description>这里的 mycluster为HA集群的逻辑名，

与hdfs-site.xml中的dfs.nameservices配置一致</description>

</property>

<name>hadoop.tmp.dir</name>

<value>/bigdata/hadoop/temp</value>

<description>这里的路径默认是NameNode、DataNode、JournalNode等存放数据的公共目录。

用户也可单独指定每类数据的存储目录。这里目录结构需要自己先创建好</description>

</property>

<name>ha.zookeeper.quorum</name>

<value>zk1.hadoop:2181,zk2.hadoop:2181,zk3.hadoop:2181</value>

<description>这里是zk集群配置中各节点的地址和端口。

注意：数量一定是奇数而且和zoo.cfg中配置的一致</description>

</property>

</configuration>

3.3、修改 $HADOOP_HOME/etc/hadoop/hfds-site.xml

100

101

<name>dfs.replication</name>

<description>配置副本数量</description>

</property>

<name>dfs.namenode.name.dir</name>

<value>file:/bigdata/hadoop/dfs/name</value>

<description>namenode元数据存储目录</description>

</property>

<name>dfs.datanode.data.dir</name>

<value>file:/bigdata/hadoop/dfs/data</value>

<description>datanode元数据存储目录</description>

</property>

<name>dfs.nameservices</name>

<value>mycluster</value>

<description>指定HA命名服务，可随意起名，

core-site.xml中fs.defaultFS配置需要引用它</description>

</property>

<name>dfs.ha.namenodes.mycluster</name>

<description>指定集群下NameNode逻辑名</description>

</property>

<name>dfs.namenode.rpc-address.mycluster.nn1</name>

<value>nn1.hadoop:9000</value>

</property>

<name>dfs.namenode.rpc-address.mycluster.nn2</name>

<value>nn2.hadoop:9000</value>

</property>

<name>dfs.namenode.http-address.mycluster.nn1</name>

<value>nn1.hadoop:50070</value>

</property>

<name>dfs.namenode.http-address.mycluster.nn2</name>

<value>nn2.hadoop:50070</value>

</property>

<name>dfs.namenode.servicerpc-address.mycluster.nn1</name>

<value>nn1.hadoop:53310</value>

</property>

<name>dfs.namenode.servicerpc-address.mycluster.nn2</name>

<value>nn2.hadoop:53310</value>

</property>

<name>dfs.ha.automatic-failover.enabled.mycluster</name>

<description>故障失败是否自动切换</description>

</property>

<name>dfs.namenode.shared.edits.dir</name>

<value>qjournal://nn1.hadoop:8485;nn2.hadoop:8485;dn1.hadoop:8485/hadoop-journal</value>

<description>配置JournalNode，包含三部分：

1.qjournal 前缀表名协议；

2.然后就是三台部署JournalNode的主机host/ip:端口，三台机器之间用分号分隔；

3.最后的hadoop-journal是journalnode的命名空间，可以随意取名。

</description>

</property>

<name>dfs.journalnode.edits.dir</name>

<value>/bigdata/hadoop/dfs/journal/</value>

<description>journalnode的本地数据存放目录</description>

</property>

<name>dfs.client.failover.proxy.provider.mycluster</name>

<value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>

<description> 指定mycluster出故障时执行故障切换的类</description>

</property>

<name>dfs.ha.fencing.methods</name>

<value>sshfence</value>

<description>ssh的操作方式执行故障切换</description>

</property>

<name>dfs.ha.fencing.ssh.private-key-files</name>

<value>/home/hadoop/.ssh/id_dsa</value>

<description> 如果使用ssh进行故障切换，使用ssh通信时用的密钥存储的位置</description>

</property>

<name>dfs.ha.fencing.ssh.connect-timeout</name>

</property>

<name>dfs.namenode.handler.count</name>

</property>

3.4、修改 $HADOOP_HOME/etc/hadoop/yarn-site.xml

<name>yarn.nodemanager.aux-services</name>

<value>mapreduce_shuffle</value>

</property>

<name>yarn.nodemanager.aux-services.mapreduce.shuffle.class</name>

<value>org.apache.hadoop.mapred.ShuffleHandler</value>

</property>

<name>yarn.resourcemanager.ha.enabled</name>

</property>

<name>yarn.resourcemanager.cluster-id</name>

<value>clusterrm</value>

</property>

<name>yarn.resourcemanager.ha.rm-ids</name>

</property>

<name>yarn.resourcemanager.hostname.rm1</name>

<value>nn1.hadoop</value>

</property>

<name>yarn.resourcemanager.hostname.rm2</name>

<value>nn2.hadoop</value>

</property>

<name>yarn.resourcemanager.recovery.enabled</name>

</property>

<name>yarn.resourcemanager.store.class</name>

<value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value>

</property>

<name>yarn.resourcemanager.zk-address</name>

<value>zk1.hadoop:2181,zk2.hadoop:2181,zk3.hadoop:2181</value>

</property>

PS: yarn-site.xml中的HA相关配置格式和hdfs-site.xml中的HA配置类似。

3.5、修改 $HADOOP_HOME/etc/hadoop/mapred-site.xml

<name>mapreduce.framework.name</name>

</property>

3.6、修改 $HADOOP_HOME/etc/hadoop/salves

dn1.hadoop

dn2.hadoop

dn3.hadoop

4、启动步骤和详细过程：

4.1、启动ZK
在所有的ZK节点执行命令： zkServer.sh start

可借助命令 zkServer.sh status 查看各个ZK的从属关系

4.2、格式化ZK（仅第一次需要做）
任意ZK节点上执行：hdfs zkfc -formatZK

[hadoop@nn1 micmiu]$ hdfs zkfc -formatZK

15/02/02 16:54:24 INFO tools.DFSZKFailoverController: Failover controller configured for NameNode NameNode at nn1.hadoop/172.17.225.61:53310

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:zookeeper.version=3.4.6-1569965, built on 02/20/2014 09:09 GMT

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:host.name=nn1.hadoop

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.version=1.7.0_75

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.vendor=Oracle Corporation

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.home=/usr/java/jdk1.7.0_75/jre

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.class.path=/usr/local/share/hadoop-2.6.0/etc/hadoop:/usr/local/share/hadoop-2.6.0/......

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.library.path=/usr/local/share/hadoop-2.6.0/lib/native

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.io.tmpdir=/tmp

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:java.compiler=<NA>

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:os.name=Linux

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:os.arch=amd64

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:os.version=2.6.32-279.el6.x86_64

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:user.name=hadoop

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:user.home=/home/hadoop

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Client environment:user.dir=/usr/local/share/hadoop-2.6.0

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Initiating client connection, connectString=zk1.hadoop:2181,zk2.hadoop:2181,zk3.hadoop:2181 sessionTimeout=5000 watcher=org.apache.hadoop.ha.ActiveStandbyElector$WatcherWithClientRef@1e884ca9

15/02/02 16:54:24 INFO zookeeper.ClientCnxn: Opening socket connection to server nn1.hadoop/172.17.225.61:2181. Will not attempt to authenticate using SASL (unknown error)

15/02/02 16:54:24 INFO zookeeper.ClientCnxn: Socket connection established to nn1.hadoop/172.17.225.61:2181, initiating session

15/02/02 16:54:24 INFO zookeeper.ClientCnxn: Session establishment complete on server nn1.hadoop/172.17.225.61:2181, sessionid = 0x14b496d55810000, negotiated timeout = 5000

15/02/02 16:54:24 INFO ha.ActiveStandbyElector: Session connected.

15/02/02 16:54:24 INFO ha.ActiveStandbyElector: Successfully created /hadoop-ha/mycluster in ZK.

15/02/02 16:54:24 INFO zookeeper.ZooKeeper: Session: 0x14b496d55810000 closed

15/02/02 16:54:24 INFO zookeeper.ClientCnxn: EventThread shut down

4.3、启动ZKFC

ZKFC（zookeeperFailoverController）是用来监控NN状态，协助实现主备NN切换的，所以仅仅在主备NN节点上启动就行。

[hadoop@nn1 micmiu]$ hadoop-daemon.sh start zkfc

starting zkfc, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-zkfc-nn1.hadoop.out

[hadoop@nn1 micmiu]$

[hadoop@nn2 micmiu]$ hadoop-daemon.sh start zkfc

starting zkfc, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-zkfc-nn2.hadoop.out

[hadoop@nn2 micmiu]$ jps

4.4、启动JournalNode 用于主备NN之间同步元数据信息的共享存储系统，在每个JN节点上启动：

#JN节点1

[hadoop@nn1 micmiu]$ hadoop-daemon.sh start journalnode

starting journalnode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-journalnode-nn1.hadoop.out

[hadoop@nn1 micmiu]$ jps

8499 QuorumPeerMain

8771 DFSZKFailoverController

8895 Jps

8837 JournalNode

#JN节点2

[hadoop@nn2 micmiu]$ hadoop-daemon.sh start journalnode

starting journalnode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-journalnode-nn2.hadoop.out

[hadoop@nn2 micmiu]$ jps

7828 QuorumPeerMain

8198 JournalNode

8082 DFSZKFailoverController

8252 Jps

#JN节点3

[hadoop@dn1 micmiu]$ hadoop-daemon.sh start journalnode

starting journalnode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-journalnode-dn1.hadoop.out

[hadoop@dn1 ~]$ jps

748 QuorumPeerMain

1008 JournalNode

1063 Jps

4.5、格式化并启动主NN
格式化：hdfs namenode -format
注意：只有第一次启动系统时需格式化，请勿重复格式化！

[hadoop@nn1 micmiu]$ hdfs namenode -format

15/02/02 17:03:05 INFO namenode.NameNode: STARTUP_MSG:

/************************************************************

STARTUP_MSG: Starting NameNode

STARTUP_MSG: host = nn1.hadoop/172.17.225.61

STARTUP_MSG: args = [-format]

STARTUP_MSG: version = 2.6.0

STARTUP_MSG: classpath = /usr/local/share/hadoop-2.6.0/etc/hadoop:/usr/local/share/hadoop-2.6.0/share/hadoop/common/lib/.......

STARTUP_MSG: build = Unknown -r Unknown; compiled by 'hadoop' on 2015-01-29T15:07Z

STARTUP_MSG: java = 1.7.0_75

************************************************************/

15/02/02 17:03:05 INFO namenode.NameNode: registered UNIX signal handlers for [TERM, HUP, INT]

15/02/02 17:03:05 INFO namenode.NameNode: createNameNode [-format]

Formatting using clusterid: CID-237f7c54-db75-470c-8baf-d4dcfaddaf2f

15/02/02 17:03:05 INFO namenode.FSNamesystem: No KeyProvider found.

15/02/02 17:03:05 INFO namenode.FSNamesystem: fsLock is fair:true

15/02/02 17:03:05 INFO blockmanagement.DatanodeManager: dfs.block.invalidate.limit=1000

15/02/02 17:03:05 INFO blockmanagement.DatanodeManager: dfs.namenode.datanode.registration.ip-hostname-check=true

15/02/02 17:03:05 INFO blockmanagement.BlockManager: dfs.namenode.startup.delay.block.deletion.sec is set to 000:00:00:00.000

15/02/02 17:03:05 INFO blockmanagement.BlockManager: The block deletion will start around 2015 Feb 02 17:03:05

15/02/02 17:03:05 INFO util.GSet: Computing capacity for map BlocksMap

15/02/02 17:03:05 INFO util.GSet: VM type = 64-bit

15/02/02 17:03:05 INFO util.GSet: 2.0% max memory 889 MB = 17.8 MB

15/02/02 17:03:05 INFO util.GSet: capacity = 2^21 = 2097152 entries

15/02/02 17:03:05 INFO blockmanagement.BlockManager: dfs.block.access.token.enable=false

15/02/02 17:03:05 INFO blockmanagement.BlockManager: defaultReplication = 3

15/02/02 17:03:05 INFO blockmanagement.BlockManager: maxReplication = 512

15/02/02 17:03:05 INFO blockmanagement.BlockManager: minReplication = 1

15/02/02 17:03:05 INFO blockmanagement.BlockManager: maxReplicationStreams = 2

15/02/02 17:03:05 INFO blockmanagement.BlockManager: shouldCheckForEnoughRacks = false

15/02/02 17:03:05 INFO blockmanagement.BlockManager: replicationRecheckInterval = 3000

15/02/02 17:03:05 INFO blockmanagement.BlockManager: encryptDataTransfer = false

15/02/02 17:03:05 INFO blockmanagement.BlockManager: maxNumBlocksToLog = 1000

15/02/02 17:03:05 INFO namenode.FSNamesystem: fsOwner = hadoop (auth:SIMPLE)

15/02/02 17:03:05 INFO namenode.FSNamesystem: supergroup = supergroup

15/02/02 17:03:05 INFO namenode.FSNamesystem: isPermissionEnabled = true

15/02/02 17:03:05 INFO namenode.FSNamesystem: Determined nameservice ID: mycluster

15/02/02 17:03:05 INFO namenode.FSNamesystem: HA Enabled: true

15/02/02 17:03:05 INFO namenode.FSNamesystem: Append Enabled: true

15/02/02 17:03:06 INFO util.GSet: Computing capacity for map INodeMap

15/02/02 17:03:06 INFO util.GSet: VM type = 64-bit

15/02/02 17:03:06 INFO util.GSet: 1.0% max memory 889 MB = 8.9 MB

15/02/02 17:03:06 INFO util.GSet: capacity = 2^20 = 1048576 entries

15/02/02 17:03:06 INFO namenode.NameNode: Caching file names occuring more than 10 times

15/02/02 17:03:06 INFO util.GSet: Computing capacity for map cachedBlocks

15/02/02 17:03:06 INFO util.GSet: VM type = 64-bit

15/02/02 17:03:06 INFO util.GSet: 0.25% max memory 889 MB = 2.2 MB

15/02/02 17:03:06 INFO util.GSet: capacity = 2^18 = 262144 entries

15/02/02 17:03:06 INFO namenode.FSNamesystem: dfs.namenode.safemode.threshold-pct = 0.9990000128746033

15/02/02 17:03:06 INFO namenode.FSNamesystem: dfs.namenode.safemode.min.datanodes = 0

15/02/02 17:03:06 INFO namenode.FSNamesystem: dfs.namenode.safemode.extension = 30000

15/02/02 17:03:06 INFO namenode.FSNamesystem: Retry cache on namenode is enabled

15/02/02 17:03:06 INFO namenode.FSNamesystem: Retry cache will use 0.03 of total heap and retry cache entry expiry time is 600000 millis

15/02/02 17:03:06 INFO util.GSet: Computing capacity for map NameNodeRetryCache

15/02/02 17:03:06 INFO util.GSet: VM type = 64-bit

15/02/02 17:03:06 INFO util.GSet: 0.029999999329447746% max memory 889 MB = 273.1 KB

15/02/02 17:03:06 INFO util.GSet: capacity = 2^15 = 32768 entries

15/02/02 17:03:06 INFO namenode.NNConf: ACLs enabled? false

15/02/02 17:03:06 INFO namenode.NNConf: XAttrs enabled? true

15/02/02 17:03:06 INFO namenode.NNConf: Maximum size of an xattr: 16384

15/02/02 17:03:07 INFO namenode.FSImage: Allocated new BlockPoolId: BP-711086735-172.17.225.61-1422867787014

15/02/02 17:03:07 INFO common.Storage: Storage directory /bigdata/hadoop/dfs/name has been successfully formatted.

15/02/02 17:03:07 INFO namenode.NNStorageRetentionManager: Going to retain 1 images with txid >= 0

15/02/02 17:03:07 INFO util.ExitUtil: Exiting with status 0

15/02/02 17:03:07 INFO namenode.NameNode: SHUTDOWN_MSG:

/************************************************************

SHUTDOWN_MSG: Shutting down NameNode at nn1.hadoop/172.17.225.61

************************************************************/

[hadoop@nn1 micmiu]$

在主NN节点执行命令启动NN: hadoop-daemon.sh start namenode

可以对比查看启动前后NN节点的进程：

#启动前

[hadoop@nn1 micmiu]$ jps

8499 QuorumPeerMain

8771 DFSZKFailoverController

8988 Jps

8837 JournalNode

[hadoop@nn1 micmiu]$ hadoop-daemon.sh start namenode

starting namenode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-namenode-nn1.hadoop.out

#启动后

[hadoop@nn1 micmiu]$ jps

8499 QuorumPeerMain

9134 Jps

8771 DFSZKFailoverController

8837 JournalNode

9017 NameNode

4.6、在备NN上同步主NN的元数据信息
hdfs namenode -bootstrapStandby

[hadoop@nn2 ~]$ hdfs namenode -bootstrapStandby

15/02/02 17:04:43 INFO namenode.NameNode: STARTUP_MSG:

/************************************************************

STARTUP_MSG: Starting NameNode

STARTUP_MSG: host = nn2.hadoop/172.17.225.121

STARTUP_MSG: args = [-bootstrapStandby]

STARTUP_MSG: version = 2.6.0

STARTUP_MSG: classpath = /usr/local/share/hadoop-2.6.0/etc/hadoop:/usr/local/share/hadoop-2.6.0......

STARTUP_MSG: build = Unknown -r Unknown; compiled by 'hadoop' on 2015-01-29T15:07Z

STARTUP_MSG: java = 1.7.0_75

************************************************************/

15/02/02 17:04:43 INFO namenode.NameNode: registered UNIX signal handlers for [TERM, HUP, INT]

15/02/02 17:04:43 INFO namenode.NameNode: createNameNode [-bootstrapStandby]

=====================================================

About to bootstrap Standby ID nn2 from:

Nameservice ID: mycluster

Other Namenode ID: nn1

Other NN's HTTP address: http://nn1.hadoop:50070

Other NN's IPC address: nn1.hadoop/172.17.225.61:53310

Namespace ID: 263802668

Block pool ID: BP-711086735-172.17.225.61-1422867787014

Cluster ID: CID-237f7c54-db75-470c-8baf-d4dcfaddaf2f

Layout version: -60

=====================================================

15/02/02 17:04:44 INFO common.Storage: Storage directory /bigdata/hadoop/dfs/name has been successfully formatted.

15/02/02 17:04:45 INFO namenode.TransferFsImage: Opening connection to http://nn1.hadoop:50070/imagetransfer?getimage=1&txid=0&storageInfo=-60:263802668:0:CID-237f7c54-db75-470c-8baf-d4dcfaddaf2f

15/02/02 17:04:45 INFO namenode.TransferFsImage: Image Transfer timeout configured to 60000 milliseconds

15/02/02 17:04:45 INFO namenode.TransferFsImage: Transfer took 0.00s at 0.00 KB/s

15/02/02 17:04:45 INFO namenode.TransferFsImage: Downloaded file fsimage.ckpt_0000000000000000000 size 352 bytes.

15/02/02 17:04:45 INFO util.ExitUtil: Exiting with status 0

15/02/02 17:04:45 INFO namenode.NameNode: SHUTDOWN_MSG:

/************************************************************

SHUTDOWN_MSG: Shutting down NameNode at nn2.hadoop/172.17.225.121

************************************************************/

[hadoop@nn2 ~]$

4.7、启动备NN

在备NN上执行命令：hadoop-daemon.sh start namenode

hadoop@nn2 ~]$ hadoop-daemon.sh start namenode

starting namenode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-namenode-nn2.hadoop.out

[hadoop@nn2 ~]$ jps

7828 QuorumPeerMain

8198 JournalNode

8082 DFSZKFailoverController

8394 NameNode

8491 Jps

4.8、设置和确认主NN

本文配置的是自动切换，ZK已经自动选择一个节点作为主NN了，所以这一步可以省略，查看节点状态：

[hadoop@nn1 ~]$ hdfs haadmin -getServiceState nn1

active

[hadoop@nn1 ~]$ hdfs haadmin -getServiceState nn2

standby

如果是配置手动切换NN的，这一步是不可缺少的，因为系统还不知道谁是主NN，两个节点的NN都是Standby状态。手动激活主NN的命令：hdfs haadmin -transitionToActive nn1

4.9、在主NN上启动Datanode

启动所有datanode命令：hadoop-daemons.sh start datanode

注意：hadoop-daemons.sh 和 hadoop-daemon.sh 命令的差异

hadoop@nn1 ~]$ hadoop-daemons.sh start datanode

dn3.hadoop: starting datanode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-datanode-dn3.hadoop.out

dn1.hadoop: starting datanode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-datanode-dn1.hadoop.out

dn2.hadoop: starting datanode, logging to /usr/local/share/hadoop-2.6.0/logs/hadoop-hadoop-datanode-dn2.hadoop.out

[hadoop@nn1 ~]$

4.10、启动YARN

方法一：一次性启动ResourceManager和NodeManager命令：start-yarn.sh

方法二：分别启动ResourceManager和NodeManager：

yarn-daemon.sh start resourcemanager
yarn-daemon.sh start nodemanager（如果有多个datanode，需使用yarn-daemons.sh）

ResourceManager 也配置了HA，根据命令查看节点状态：

yarn rmadmin –getServiceState serviceid

[hadoop@nn1 ~]$ yarn rmadmin -getServiceState rm1

active

[hadoop@nn1 ~]$ yarn rmadmin -getServiceState rm2

standby

[hadoop@nn1 ~]$

4.11 启动MR JobHistory Server

在dn1.hadoop上运行MRJS ：mr-jobhistory-daemon.sh start historyserver

//运行MRJS之前

[hadoop@dn1 ~]$ jps

14625 Jps

3568 NodeManager

748 QuorumPeerMain

1008 JournalNode

1194 DataNode

[hadoop@dn1 ~]$ mr-jobhistory-daemon.sh start historyserver

starting historyserver, logging to /usr/local/share/hadoop-2.6.0/logs/mapred-hadoop-historyserver-dn1.hadoop.out

//运行MRJS之后

[hadoop@dn1 ~]$ jps

14745 JobHistoryServer

3568 NodeManager

748 QuorumPeerMain

1008 JournalNode

1194 DataNode

14786 Jps

4.12、验证NameNode 和ResourceManager 的HA是否生效

把当前主节点中的相关进程kill掉查看各节点状态切换情况。

4.13、验证NN HA的透明性
注意验证 hdfs dfs -ls / 和 hdfs dfs -ls hdfs://mycluster/ 的访问效果是一致的：

[hadoop@nn1 ~]$ hdfs dfs -ls /

Found 2 items

drwx------ - hadoop supergroup 0 2015-02-02 23:42 /tmp

drwxr-xr-x - hadoop supergroup 0 2015-02-02 23:39 /user

[hadoop@nn1 ~]$ hdfs dfs -ls hdfs://mycluster/

Found 2 items

drwx------ - hadoop supergroup 0 2015-02-02 23:42 hdfs://mycluster/tmp

drwxr-xr-x - hadoop supergroup 0 2015-02-02 23:39 hdfs://mycluster/user

[hadoop@nn1 ~]$

[五]、运行wrodcount demo

这个demo的演示可参考：hadoop 2.2.0 集群模式安装配置和测试中的 wordcount演示步骤，这里不再重复描述了。

(责任编辑：IT)