Hadoop集群(二) HDFS搭建

HDFS只是Hadoop最基本的一个服务，很多其他服务，都是基于HDFS展开的。所以部署一个HDFS集群，是很核心的一个动作，也是大数据平台的开始。

   安装Hadoop集群，首先需要有Zookeeper才可以完成安装。如果没有Zookeeper，请先部署一套Zookeeper。另外，JDK以及物理主机的一些设置等。请参考：

Hadoop集群(一) Zookeeper搭建

Hadoop集群(三) Hbase搭建

Hadoop集群(四) Hadoop升级

下面开始HDFS的安装

HDFS主机分配

1 2 3

192.168.67.101 c6701 --Namenode+datanode 192.168.67.102 c6702 --datanode 192.168.67.103 c6703 --datanode

1. 安装HDFS，解压hadoop-2.6.0-EDH-0u2.tar.gz 

我同时下载2.6和2.7版本的软件，先安装2.6，然后在执行2.6到2.7的升级步骤

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

useradd hdfs echo "hdfs:hdfs" | chpasswd su - hdfs cd /tmp/software tar -zxvf hadoop-2.6.0-EDH-0u2.tar.gz -C /home/hdfs/ mkdir -p /data/hadoop/temp  mkdir -p /data/hadoop/journal  mkdir -p /data/hadoop/hdfs/name  mkdir -p /data/hadoop/hdfs/data chown -R hdfs:hdfs /data/hadoop chown -R hdfs:hdfs /data/hadoop/temp  chown -R hdfs:hdfs /data/hadoop/journal  chown -R hdfs:hdfs /data/hadoop/hdfs/name  chown -R hdfs:hdfs /data/hadoop/hdfs/data  $ pwd /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop

2. 修改core-site.xml对应的参数

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

$ cat core-site.xml <configuration>  <!-- 指定hdfs的nameservice为ns -->  <property>          <name>fs.defaultFS</name>          <value>hdfs://ns</value>     </property>  <!--指定hadoop数据临时存放目录-->  <property>       <name>hadoop.tmp.dir</name>       <value>/data/hadoop/temp</value>  </property>                              <property>          <name>io.file.buffer.size</name>          <value>4096</value>     </property>  <!--指定zookeeper地址-->  <property>       <name>ha.zookeeper.quorum</name>       <value>c6701:2181,c6702:2181,c6703:2181</value>  </property>  </configuration>

3. 修改hdfs-site.xml对应的参数

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83

cat hdfs-site.xml <configuration>     <!--指定hdfs的nameservice为ns，需要和core-site.xml中的保持一致,并且ns如果改，整个文件中，全部的ns要都修改，保持统一 -->        <property>            <name>dfs.nameservices</name>            <value>ns</value>        </property>      <!-- ns下面有两个NameNode，分别是nn1，nn2 -->     <property>       <name>dfs.ha.namenodes.ns</name>       <value>nn1,nn2</value>     </property>     <!-- nn1的RPC通信地址 -->     <property>       <name>dfs.namenode.rpc-address.ns.nn1</name>       <value>c6701:9000</value>     </property>     <!-- nn1的http通信地址 -->     <property>         <name>dfs.namenode.http-address.ns.nn1</name>         <value>c6701:50070</value>     </property>     <!-- nn2的RPC通信地址 -->     <property>         <name>dfs.namenode.rpc-address.ns.nn2</name>         <value>c6702:9000</value>     </property>     <!-- nn2的http通信地址 -->     <property>         <name>dfs.namenode.http-address.ns.nn2</name>         <value>c6702:50070</value>     </property>     <!-- 指定NameNode的元数据在JournalNode上的存放位置 -->     <property>         <name>dfs.namenode.shared.edits.dir</name>         <value>qjournal://c6701:8485;c6702:8485;c6703:8485/ns</value>     </property>     <!-- 指定JournalNode在本地磁盘存放数据的位置 -->     <property>           <name>dfs.journalnode.edits.dir</name>           <value>/data/hadoop/journal</value>     </property>     <!-- 开启NameNode故障时自动切换 -->     <property>           <name>dfs.ha.automatic-failover.enabled</name>           <value>true</value>     </property>     <!-- 配置失败自动切换实现方式 -->     <property>             <name>dfs.client.failover.proxy.provider.ns</name>             <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>     </property>     <!-- 配置隔离机制 -->     <property>             <name>dfs.ha.fencing.methods</name>             <value>sshfence</value>     </property>     <!-- 使用隔离机制时需要ssh免登陆 -->     <property>             <name>dfs.ha.fencing.ssh.private-key-files</name>             <value>/home/hdfs/.ssh/id_rsa</value>     </property>                                    <property>            <name>dfs.namenode.name.dir</name>            <value>/data/hadoop/hdfs/name</value>        </property>             <property>            <name>dfs.datanode.data.dir</name>            <value>/data/hadoop/hdfs/data</value>        </property>             <property>          <name>dfs.replication</name>          <value>2</value>        </property>      <!-- 在NN和DN上开启WebHDFS (REST API)功能,不是必须 -->                                                                        <property>          <name>dfs.webhdfs.enabled</name>          <value>true</value>        </property>    </configuration>

4. 添加slaves文件

1 2 3 4

$ more slaves c6701 c6702 c6703

--- 安装C6702的hdfs---

5. 创建c6702的用户，并为hdfs用户ssh免密

1 2 3

ssh c6702 "useradd hdfs" ssh c6702 "echo "hdfs:hdfs" | chpasswd" ssh-copy-id  hdfs@c6702

6. 拷贝软件

1 2	scp -r /tmp/software/hadoop-2.6.0-EDH-0u2.tar.gz root@c6702:/tmp/software/. ssh c6702 "chmod 777 /tmp/software/*"

7. 创建目录，解压软件

1 2 3 4

ssh hdfs@c6702 "mkdir hdfs" ssh hdfs@c6702 "tar -zxvf /tmp/software/hadoop-2.6.0-EDH-0u2.tar.gz -C /home/hdfs" ssh hdfs@c6702 "ls -al hdfs" ssh hdfs@c6702 "ls -al hdfs/hadoop*"

复制配置文件

1 2 3 4 5

ssh hdfs@c6702 "rm -rf /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml" ssh hdfs@c6702 "rm -rf /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml" scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml hdfs@c6702:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml hdfs@c6702:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/slaves hdfs@c6702:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/slaves

创建hdfs需要的目录

1 2 3 4 5 6

ssh root@c6702 "mkdir -p /data/hadoop" ssh root@c6702 " chown -R hdfs:hdfs  /data/hadoop" ssh hdfs@c6702 "mkdir -p /data/hadoop/temp" ssh hdfs@c6702 "mkdir -p /data/hadoop/journal" ssh hdfs@c6702 "mkdir -p /data/hadoop/hdfs/name" ssh hdfs@c6702 "mkdir -p /data/hadoop/hdfs/data"

--- 安装C6703的hdfs---

8. 创建c6703的用户，并为hdfs用户ssh免密

1 2 3

ssh c6703 "useradd hdfs" ssh c6703 "echo "hdfs:hdfs" | chpasswd" ssh-copy-id  hdfs@c6703

9. 拷贝软件

1 2 3 4 5 6 7

scp -r /tmp/software/hadoop-2.6.0-EDH-0u2.tar.gz root@c6703:/tmp/software/. ssh c6703 "chmod 777 /tmp/software/*" 10. 创建目录，解压软件 ssh hdfs@c6703 "mkdir hdfs" ssh hdfs@c6703 "tar -zxvf /tmp/software/hadoop-2.6.0-EDH-0u2.tar.gz -C /home/hdfs" ssh hdfs@c6703 "ls -al hdfs" ssh hdfs@c6703 "ls -al hdfs/hadoop*"

复制配置文件

1 2 3 4 5

ssh hdfs@c6703 "rm -rf /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml" ssh hdfs@c6703 "rm -rf /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml" scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml hdfs@c6703:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/core-site.xml scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml hdfs@c6703:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/hdfs-site.xml scp -r /home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/slaves hdfs@c6703:/home/hdfs/hadoop-2.6.0-EDH-0u2/etc/hadoop/slaves

创建hdfs需要的目录

1 2 3 4 5 6

ssh root@c6703 "mkdir -p /data/hadoop" ssh root@c6703 " chown -R hdfs:hdfs  /data/hadoop" ssh hdfs@c6703 "mkdir -p /data/hadoop/temp" ssh hdfs@c6703 "mkdir -p /data/hadoop/journal" ssh hdfs@c6703 "mkdir -p /data/hadoop/hdfs/name" ssh hdfs@c6703 "mkdir -p /data/hadoop/hdfs/data"

11. 启动HDFS，先启动三个节点的journalnode

/home/hdfs/hadoop-2.6.0-EDH-0u2/sbin/hadoop-daemon.sh start journalnode

检查状态

1 2 3

$ jps 3958 Jps 3868 JournalNode

12. 然后启动namenode,首次启动namenode之前，先在其中一个节点(主节点)format namenode信息，信息会存在于dfs.namenode.name.dir指定的路径中

1 2	<name>dfs.namenode.name.dir</name>     <value>/data/hadoop/hdfs/name</value>

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

$ ./hdfs namenode -format 17/09/26 07:52:17 INFO namenode.NameNode: STARTUP_MSG: /************************************************************ STARTUP_MSG: Starting NameNode STARTUP_MSG:   host = c6701.python279.org/192.168.67.101 STARTUP_MSG:   args = [-format] STARTUP_MSG:   version = 2.6.0-EDH-0u2 STARTUP_MSG:   classpath = /home/hdfs/hadoop-2.6.0-EDHxxxxxxxxxx STARTUP_MSG:   build = http://gitlab-xxxxx STARTUP_MSG:   java = 1.8.0_144 ************************************************************/ 17/09/26 07:52:17 INFO namenode.NameNode: registered UNIX signal handlers for [TERM, HUP, INT] 17/09/26 07:52:17 INFO namenode.NameNode: createNameNode [-format] 17/09/26 07:52:18 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. 17/09/26 07:52:18 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. Formatting using clusterid: CID-b2f01411-862f-44b2-a6dc-7d17bd48d522 17/09/26 07:52:18 INFO namenode.FSNamesystem: No KeyProvider found. 17/09/26 07:52:18 INFO namenode.FSNamesystem: fsLock is fair:true 17/09/26 07:52:18 INFO blockmanagement.DatanodeManager: dfs.block.invalidate.limit=1000 17/09/26 07:52:18 INFO blockmanagement.DatanodeManager: dfs.namenode.datanode.registration.ip-hostname-check=true 17/09/26 07:52:18 INFO blockmanagement.BlockManager: dfs.namenode.startup.delay.block.deletion.sec is set to 000:00:00:00.000 17/09/26 07:52:18 INFO blockmanagement.BlockManager: The block deletion will start around 2017 Sep 26 07:52:18 17/09/26 07:52:18 INFO util.GSet: Computing capacity for map BlocksMap 17/09/26 07:52:18 INFO util.GSet: VM type       = 64-bit 17/09/26 07:52:18 INFO util.GSet: 2.0% max memory 966.7 MB = 19.3 MB 17/09/26 07:52:18 INFO util.GSet: capacity      = 2^21 = 2097152 entries 17/09/26 07:52:18 INFO blockmanagement.BlockManager: dfs.block.access.token.enable=false 17/09/26 07:52:18 INFO blockmanagement.BlockManager: defaultReplication         = 2 17/09/26 07:52:18 INFO blockmanagement.BlockManager: maxReplication             = 512 17/09/26 07:52:18 INFO blockmanagement.BlockManager: minReplication             = 1 17/09/26 07:52:18 INFO blockmanagement.BlockManager: maxReplicationStreams      = 2 17/09/26 07:52:18 INFO blockmanagement.BlockManager: shouldCheckForEnoughRacks  = false 17/09/26 07:52:18 INFO blockmanagement.BlockManager: replicationRecheckInterval = 3000 17/09/26 07:52:18 INFO blockmanagement.BlockManager: encryptDataTransfer        = false 17/09/26 07:52:18 INFO blockmanagement.BlockManager: maxNumBlocksToLog          = 1000 17/09/26 07:52:18 INFO namenode.FSNamesystem: fsOwner             = hdfs (auth:SIMPLE) 17/09/26 07:52:18 INFO namenode.FSNamesystem: supergroup          = supergroup 17/09/26 07:52:18 INFO namenode.FSNamesystem: isPermissionEnabled = true 17/09/26 07:52:18 INFO namenode.FSNamesystem: Determined nameservice ID: ns 17/09/26 07:52:18 INFO namenode.FSNamesystem: HA Enabled: true 17/09/26 07:52:18 INFO namenode.FSNamesystem: Append Enabled: true 17/09/26 07:52:18 INFO util.GSet: Computing capacity for map INodeMap 17/09/26 07:52:18 INFO util.GSet: VM type       = 64-bit 17/09/26 07:52:18 INFO util.GSet: 1.0% max memory 966.7 MB = 9.7 MB 17/09/26 07:52:18 INFO util.GSet: capacity      = 2^20 = 1048576 entries 17/09/26 07:52:18 INFO namenode.NameNode: Caching file names occuring more than 10 times 17/09/26 07:52:18 INFO util.GSet: Computing capacity for map cachedBlocks 17/09/26 07:52:18 INFO util.GSet: VM type       = 64-bit 17/09/26 07:52:18 INFO util.GSet: 0.25% max memory 966.7 MB = 2.4 MB 17/09/26 07:52:18 INFO util.GSet: capacity      = 2^18 = 262144 entries 17/09/26 07:52:18 INFO namenode.FSNamesystem: dfs.namenode.safemode.threshold-pct = 0.9990000128746033 17/09/26 07:52:18 INFO namenode.FSNamesystem: dfs.namenode.safemode.min.datanodes = 0 17/09/26 07:52:18 INFO namenode.FSNamesystem: dfs.namenode.safemode.extension     = 30000 17/09/26 07:52:18 INFO namenode.FSNamesystem: Retry cache on namenode is enabled 17/09/26 07:52:18 INFO namenode.FSNamesystem: Retry cache will use 0.03 of total heap and retry cache entry expiry time is 600000 millis 17/09/26 07:52:18 INFO util.GSet: Computing capacity for map NameNodeRetryCache 17/09/26 07:52:18 INFO util.GSet: VM type       = 64-bit 17/09/26 07:52:18 INFO util.GSet: 0.029999999329447746% max memory 966.7 MB = 297.0 KB 17/09/26 07:52:18 INFO util.GSet: capacity      = 2^15 = 32768 entries 17/09/26 07:52:18 INFO namenode.NNConf: ACLs enabled? false 17/09/26 07:52:18 INFO namenode.NNConf: XAttrs enabled? true 17/09/26 07:52:18 INFO namenode.NNConf: Maximum size of an xattr: 16384 17/09/26 07:52:19 INFO namenode.FSImage: Allocated new BlockPoolId: BP-144216011-192.168.67.101-1506412339757 17/09/26 07:52:19 INFO common.Storage: Storage directory /data/hadoop/hdfs/name has been successfully formatted. 17/09/26 07:52:20 INFO namenode.NNStorageRetentionManager: Going to retain 1 images with txid >= 0 17/09/26 07:52:20 INFO util.ExitUtil: Exiting with status 0 17/09/26 07:52:20 INFO namenode.NameNode: SHUTDOWN_MSG: /************************************************************ SHUTDOWN_MSG: Shutting down NameNode at c6701.python279.org/192.168.67.101 ************************************************************/

13. standby namenode需要先执行bootstrapstandby，输出如下

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

[hdfs@c6702 sbin]$ ../bin/hdfs namenode -bootstrapstandby 17/09/26 09:44:58 INFO namenode.NameNode: STARTUP_MSG: /************************************************************ STARTUP_MSG: Starting NameNode STARTUP_MSG:   host = c6702.python279.org/192.168.67.102 STARTUP_MSG:   args = [-bootstrapstandby] STARTUP_MSG:   version = 2.6.0-EDH-0u2 STARTUP_MSG:   classpath = /home/hdfs/haxxx STARTUP_MSG:   build = http://gitlab-xxxx STARTUP_MSG:   java = 1.8.0_144 ************************************************************/ 17/09/26 09:44:58 INFO namenode.NameNode: registered UNIX signal handlers for [TERM, HUP, INT] 17/09/26 09:44:58 INFO namenode.NameNode: createNameNode [-bootstrapstandby] 17/09/26 09:44:59 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. 17/09/26 09:44:59 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. ===================================================== About to bootstrap Standby ID nn2 from:            Nameservice ID: ns         Other Namenode ID: nn1   Other NN's HTTP address: http://c6701:50070   Other NN's IPC  address: c6701/192.168.67.101:9000              Namespace ID: 793662207             Block pool ID: BP-144216011-192.168.67.101-1506412339757                Cluster ID: CID-b2f01411-862f-44b2-a6dc-7d17bd48d522            Layout version: -60 ===================================================== Re-format filesystem in Storage Directory /data/hadoop/hdfs/name ? (Y or N) y 17/09/26 09:45:16 INFO common.Storage: Storage directory /data/hadoop/hdfs/name has been successfully formatted. 17/09/26 09:45:16 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. 17/09/26 09:45:16 WARN common.Util: Path /data/hadoop/hdfs/name should be specified as a URI in configuration files. Please update hdfs configuration. 17/09/26 09:45:17 INFO namenode.TransferFsImage: Opening connection to http://c6701:50070/imagetransfer?getimage=1&txid=0&storageInfo=-60:793662207:0:CID-b2f01411-862f-44b2-a6dc-7d17bd48d522 17/09/26 09:45:17 INFO namenode.TransferFsImage: Image Transfer timeout configured to 60000 milliseconds 17/09/26 09:45:17 INFO namenode.TransferFsImage: Transfer took 0.01s at 0.00 KB/s 17/09/26 09:45:17 INFO namenode.TransferFsImage: Downloaded file fsimage.ckpt_0000000000000000000 size 351 bytes. 17/09/26 09:45:17 INFO util.ExitUtil: Exiting with status 0 17/09/26 09:45:17 INFO namenode.NameNode: SHUTDOWN_MSG: /************************************************************ SHUTDOWN_MSG: Shutting down NameNode at c6702.python279.org/192.168.67.102 ************************************************************/

14. 检查状态，namenode还没有启动

1 2 3

[hdfs@c6702 sbin]$ jps 4539 Jps 3868 JournalNode

15. 启动standby namenode，命令和master启动的方式相同

1 2	[hdfs@c6702 sbin]$ ./hadoop-daemon.sh start namenode starting namenode, logging to /home/hdfs/hadoop-2.6.0-EDH-0u2/logs/hadoop-hdfs-namenode-c6702.python279.org.out

16. 再次检查，namenode已经启动

1 2 3 4

[hdfs@c6702 sbin]$ jps 4640 Jps 4570 NameNode 3868 JournalNode

17. 格式化zkfc,让在zookeeper中生成ha节点，在master上执行如下命令，完成格式化

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

[hdfs@c6701 bin]$ ./hdfs zkfc -formatZK 17/09/26 09:59:20 INFO tools.DFSZKFailoverController: Failover controller configured for NameNode NameNode at c6701/192.168.67.101:9000 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:zookeeper.version=3.4.6-1569965, built on 02/20/2014 09:09 GMT 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:host.name=c6701.python279.org 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.version=1.8.0_144 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.vendor=Oracle Corporation 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.home=/usr/local/jdk1.8.0_144/jre 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.class.path=/home/hdfs/hadoop-2.6.0-EDH-0u2/exxxx 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.library.path=/home/hdfs/hadoop-2.6.0-EDH-0u2/lib/native 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.io.tmpdir=/tmp 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:java.compiler=<NA> 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:os.name=Linux 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:os.arch=amd64 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:os.version=2.6.32-573.el6.x86_64 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:user.name=hdfs 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:user.home=/home/hdfs 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Client environment:user.dir=/home/hdfs/hadoop-2.6.0-EDH-0u2/bin 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Initiating client connection, connectString=c6701:2181,c6702:2181,c6703:2181 sessionTimeout=5000 watcher=org.apache.hadoop.ha.ActiveStandbyElector$WatcherWithClientRef@20deea7f 17/09/26 09:59:20 INFO zookeeper.ClientCnxn: Opening socket connection to server c6703.python279.org/192.168.67.103:2181. Will not attempt to authenticate using SASL (unknown error) 17/09/26 09:59:20 INFO zookeeper.ClientCnxn: Socket connection established to c6703.python279.org/192.168.67.103:2181, initiating session 17/09/26 09:59:20 INFO zookeeper.ClientCnxn: Session establishment complete on server c6703.python279.org/192.168.67.103:2181, sessionid = 0x35ebc5163710000, negotiated timeout = 5000 17/09/26 09:59:20 INFO ha.ActiveStandbyElector: Session connected. 17/09/26 09:59:20 INFO ha.ActiveStandbyElector: Successfully created /hadoop-ha/ns in ZK. 17/09/26 09:59:20 INFO zookeeper.ZooKeeper: Session: 0x35ebc5163710000 closed 17/09/26 09:59:20 INFO zookeeper.ClientCnxn: EventThread shut down

18. 格式化完成的检查

格式成功后，查看zookeeper中可以看到    <<<<<<<<<<<命令没确认

1	[zk: localhost:2181(CONNECTED) 1] ls /hadoop-ha

19. 启动zkfc，这个就是为namenode使用的

1 2 3 4 5 6 7 8

./hadoop-daemon.sh start zkfc starting zkfc, logging to /home/hdfs/hadoop-2.6.0-EDH-0u2/logs/hadoop-hdfs-zkfc-c6701.python279.org.out $ jps 4272 DataNode 4402 JournalNode 6339 Jps 6277 DFSZKFailoverController 4952 NameNode

20. 另一个节点启动zkfc，

1 2 3 4 5 6 7

ssh  hdfs@c6702  /home/hdfs/hadoop-2.6.0-EDH-0u2/sbin/hadoop-daemon.sh start zkfc $ jps 4981 Jps 4935 DFSZKFailoverController 4570 NameNode 3868 JournalNode

21. 注意：进行初始化的时候，必须保证zk集群已经启动了。

    1、在ZK中创建znode来存储automatic Failover的数据，任选一个NN执行完成即可：

        sh bin/hdfs zkfc -formatZK

    2、启动zkfs，在所有的NN节点中执行以下命令：

        sh sbin/hadoop-daemon.sh start zkfc

22. 启动datanode

最后启动集群

1 2	/home/hdfs/hadoop-2.6.0-EDH-0u2/sbin/hadoop-daemon.sh start zkfc     sh sbin/start-dfs.sh

HDFS安装过程中的重点，最后在软件启动过程中，一些初始化操作，很重要。

1. 启动全部的journalnode

2. 在namenode1上执行， hdfs namenode -format

3. 在namenode1上执行， 启动namenode1，命令hadoop-daemon.sh start namenode 

4. 在namenode2上执行, hdfs namenode -bootstrapstandby

5. 在namenode1上执行，格式化zkfc,在zookeeper中生成HA节点, hdfs zkfc -formatZK

6. 启动zkfc，hadoop-daemon.sh start zkfc。 有namenode运行的节点，都要启动ZKFC

7. 启动 datanode

HDFS只是Hadoop最基本的一个模块，这里已经安装完成，可以为后面的Hbase提供服务了。

http://blog.51cto.com/hsbxxl/1971475