因为需要部署spark环境，特意重新安装了一个测试的hadoop集群，现将相关步骤记录如下：
创新互联专注为客户提供全方位的互联网综合服务，包含不限于做网站、
网站制作、黄南州网络推广、小程序定制开发、黄南州网络营销、黄南州企业策划、黄南州品牌公关、搜索引擎seo、人物专访、企业宣传片、企业代运营等，从售前售中售后，我们都将竭诚为您服务，您的肯定，是我们大的嘉奖；
创新互联为所有大学生创业者提供
黄南州建站搭建服务，24小时服务热线：13518219792，官方网址：www.cdcxhl.com
硬件环境：四台虚拟机，hadoop1~hadoop4，3G内存，60G硬盘，2核CPU
软件环境：CentOS6.5，hadoop-2.6.0-cdh6.8.2，JDK1.7
部署规划：
hadoop1（192.168.0.3）：namenode（active）、resourcemanager
hadoop2（192.168.0.4）：namenode（standby）、journalnode、datanode、nodemanager、historyserver
hadoop3（192.168.0.5）：journalnode、datanode、nodemanager
hadoop4（192.168.0.6）：journalnode、datanode、nodemanager
HDFS的HA采用QJM的方式（journalnode）：
一、系统准备
1、每台机关闭selinux
#vi /etc/selinux/config
SELINUX=disabled
2、每台机关闭防火墙（切记要关闭，否则格式化hdfs时会报错无法连接journalnode）
#chkconfig iptables off
#service iptables stop
3、每台机安装jdk1.7
#cd /software
#tar -zxf jdk-7u65-linux-x64.gz -C /opt/
#cd /opt
#ln -s jdk-7u65-linux-x64.gz java
#vi /etc/profile
export JAVA_HOME=/opt/java
export PATH=$PATH:$JAVA_HOME/bin
4、每台机建立hadoop相关用户，并建立互信
#useradd grid
#passwd grid
（建立互信步骤略）
5、每台机建立相关目录
#mkdir -p /hadoop_data/hdfs/name
#mkdir -p /hadoop_data/hdfs/data
#mkdir -p /hadoop_data/hdfs/journal
#mkdir -p /hadoop_data/yarn/local
#chown -R grid:grid /hadoop_data
二、hadoop部署
HDFS HA主要是指定nameservices（如果不做HDFS ferderation，就只会有一个ID），同时指定该nameserviceID下面的两个namenode及其地址。此处的nameservice名设置为hadoop-spark
1、每台机解压hadoop包
#cd /software
#tar -zxf hadoop-2.6.0-cdh6.8.2.tar.gz -C /opt/
#cd /opt
#chown -R grid:grid hadoop-2.6.0-cdh6.8.2
#ln -s hadoop-2.6.0-cdh6.8.2 hadoop
2、切换到grid用户继续操作
#su - grid
$cd /opt/hadoop/etc/hadoop
3、配置hadoop-env.sh（其实只配置JAVA_HOME）
$vi hadoop-env.sh
# The java implementation to use.
export JAVA_HOME=/opt/java
4、设置hdfs-site.xml
dfs.replication
1
dfs.nameservices
hadoop-spark
Comma-separated list of nameservices.
dfs.ha.namenodes.hadoop-spark
nn1,nn2
The prefix for a given nameservice, contains a comma-separated
list of namenodes for a given nameservice (eg EXAMPLENAMESERVICE).
dfs.namenode.rpc-address.hadoop-spark.nn1
hadoop1:8020
RPC address for nomenode1 of hadoop-spark
dfs.namenode.rpc-address.hadoop-spark.nn2
hadoop2:8020
RPC address for nomenode2 of hadoop-spark
dfs.namenode.http-address.hadoop-spark.nn1
hadoop1:50070
The address and the base port where the dfs namenode1 web ui will listen on.
dfs.namenode.http-address.hadoop-spark.nn2
hadoop2:50070
The address and the base port where the dfs namenode2 web ui will listen on.
dfs.namenode.name.dir
file:///hadoop_data/hdfs/name
Determines where on the local filesystem the DFS name node
should store the name table(fsp_w_picpath). If this is a comma-delimited list
of directories then the name table is replicated in all of the
directories, for redundancy.
dfs.namenode.shared.edits.dir
qjournal://hadoop2:8485;hadoop3:8485;hadoop4:8485/hadoop-spark
A directory on shared storage between the multiple namenodes
in an HA cluster. This directory will be written by the active and read
by the standby in order to keep the namespaces synchronized. This directory
does not need to be listed in dfs.namenode.edits.dir above. It should be
left empty in a non-HA cluster.
dfs.datanode.data.dir
file:///hadoop_data/hdfs/data
Determines where on the local filesystem an DFS data node
should store its blocks. If this is a comma-delimited
list of directories, then data will be stored in all named
directories, typically on different devices.
Directories that do not exist are ignored.
dfs.client.failover.proxy.provider.hadoop-spark
org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider
dfs.ha.automatic-failover.enabled
false
Whether automatic failover is enabled. See the HDFS High
Availability documentation for details on automatic HA
configuration.
dfs.journalnode.edits.dir
/hadoop_data/hdfs/journal
5、配置core-site.xml（配置fs.defaultFS，使用HA的nameservices名称）
fs.defaultFS
hdfs://hadoop-spark
The name of the default file system. A URI whose
scheme and authority determine the FileSystem implementation. The
uri's scheme determines the config property (fs.SCHEME.impl) naming
the FileSystem implementation class. The uri's authority is used to
determine the host, port, etc. for a filesystem.
6、配置mapred-site.xml
mapreduce.framework.name
yarn
The runtime framework for executing MapReduce jobs.
Can be one of local, classic or yarn.
mapreduce.jobhistory.address
hadoop2:10020
MapReduce JobHistory Server IPC host:port
mapreduce.jobhistory.webapp.address
hadoop2:19888
MapReduce JobHistory Server Web UI host:port
7、配置yarn-site.xml
The hostname of the RM.
yarn.resourcemanager.hostname
hadoop1
The address of the applications manager interface in the RM.
yarn.resourcemanager.address
${yarn.resourcemanager.hostname}:8032
The address of the scheduler interface.
yarn.resourcemanager.scheduler.address
${yarn.resourcemanager.hostname}:8030
The http address of the RM web application.
yarn.resourcemanager.webapp.address
${yarn.resourcemanager.hostname}:8088
The https adddress of the RM web application.
yarn.resourcemanager.webapp.https.address
${yarn.resourcemanager.hostname}:8090
yarn.resourcemanager.resource-tracker.address
${yarn.resourcemanager.hostname}:8031
The address of the RM admin interface.
yarn.resourcemanager.admin.address
${yarn.resourcemanager.hostname}:8033
The class to use as the resource scheduler.
yarn.resourcemanager.scheduler.class
org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.FairScheduler
fair-scheduler conf location
yarn.scheduler.fair.allocation.file
${yarn.home.dir}/etc/hadoop/fairscheduler.xml
List of directories to store localized files in. An
application's localized file directory will be found in:
${yarn.nodemanager.local-dirs}/usercache/${user}/appcache/application_${appid}.
Individual containers' work directories, called container_${contid}, will
be subdirectories of this.
yarn.nodemanager.local-dirs
/hadoop_data/yarn/local
Whether to enable log aggregation
yarn.log-aggregation-enable
true
Where to aggregate logs to.
yarn.nodemanager.remote-app-log-dir
/tmp/logs
Amount of physical memory, in MB, that can be allocated
for containers.
yarn.nodemanager.resource.memory-mb
2048
Number of CPU cores that can be allocated
for containers.
yarn.nodemanager.resource.cpu-vcores
2
the valid service name should only contain a-zA-Z0-9_ and can not start with numbers
yarn.nodemanager.aux-services
mapreduce_shuffle
8、配置slaves
hadoop2
hadoop3
hadoop4
9、配置fairscheduler.xml
0mb, 0 vcores
6144 mb, 6 vcores
50
300
1.0
grid
10、同步配置文件到各个节点
$cd /opt/hadoop/etc
$scp -r hadoop hadoop2:/opt/hadoop/etc/
$scp -r hadoop hadoop3:/opt/hadoop/etc/
$scp -r hadoop hadoop4:/opt/hadoop/etc/
三、启动集群（格式化文件系统）
1、建立环境变量
$vi ~/.bash_profile
export HADOOP_HOME=/opt/hadoop
export YARN_HOME_DIR=/opt/hadoop
export HADOOP_CONF_DIR=/opt/hadoop/etc/hadoop
export YARN_CONF_DIR=/opt/hadoop/etc/hadoop
2、启动HDFS
先启动journalnode，在hadoop2~hadoop4上：
$cd /opt/hadoop/
$sbin/hadoop-daemon.sh start journalnode
格式化HDFS，然后启动namenode。在hadoop1上：
$bin/hdfs namenode -format
$sbin/hadoop-daemon.sh start namenode
同步另一个namenode，并启动。在hadoop2上：
$bin/hdfs namenode -bootstrapStandby
$sbin/hadoop-daemon.sh start namenode
此时两个namenode都是standby状态，将hadoop1切换成active（hadoop1在hdfs-site.xml里对应的是nn1）：
$bin/hdfs haadmin -transitionToActive nn1
启动datanode。在hadoop1上（active的namenode）：
$sbin/hadoop-daemons.sh start datanode
注意事项：后续启动，只需使用sbin/start-dfs.sh即可。但由于没有配置zookeeper的failover，所以只能HA只能使用手工切换。所以每次启动HDFS，都要执行$bin/hdfs haadmin -transitionToActive nn1来使hadoop1的namenode变成active状态
2、启动yarn
在hadoop1上（resourcemanager）：
$sbin/start-yarn.sh
————————————————————————————————————————————
以上配置的HDFS HA并不是自动故障切换的，如果配置HDFS自动故障切换，需要添加以下步骤（先停掉集群）：
1、部署zookeeper，步骤省略。部署在hadoop2、hadoop3、hadoop4，并启动
2、在hdfs-site.xml中添加：
dfs.ha.automatic-failover.enabled
true
dfs.ha.fencing.methods
sshfence
dfs.ha.fencing.ssh.private-key-files
/home/exampleuser/.ssh/id_rsa
解释详见官方文档。这样配置设定了fencing方法是通过ssh去关闭前一个活动节点的端口。前提前两个namenode能互相SSH。
还有另外一种配置方法：
dfs.ha.automatic-failover.enabled
true
dfs.ha.fencing.methods
shell(/path/to/my/script.sh arg1 arg2 ...)
这样的配置实际上是使用shell来隔绝端口和程序。如果不想做实际的动作，dfs.ha.fencing.methods可配置成shell(/bin/true)
3、在core-site.xml中添加：
ha.zookeeper.quorum
hadoop2:2181,hadoop3:2181,hadoop4:2181
4、初始化zkfc（在namenode上执行）
bin/hdfs zkfc -formatZK
5、启动集群
___________________________________________________________________________________________________
zkfc：每个namenode上都运行，是zk的客户端，负责自动故障切换
zk：奇数个节点，维护一致性锁、负责选举活动节点
joural node：奇数个节点，用于active和standby节点之间数据同步。活动节点把数据写入这些节点，standby节点读取
————————————————————————————————————————————
更改成resourcemanager HA：
选择hadoop2作为另一个rm节点
1、设置hadoop2对其它节点作互信
2、编译yarn-site.xml并同步到其它机器
3、复制fairSheduler.xml到hadoop2
4、启动rm
5、启动另一个rm
另外有需要云服务器可以了解下创新互联scvps.cn，海内外云服务器15元起步，三天无理由+7*72小时售后在线，公司持有idc许可证，提供“云服务器、裸金属服务器、高防服务器、香港服务器、美国服务器、虚拟主机、免备案服务器”等云主机租用服务以及企业上云的综合解决方案，具有“安全稳定、简单易用、服务可用性高、性价比高”等特点与优势，专为企业上云打造定制，能够满足用户丰富、多元化的应用场景需求。
网页名称：hadoop2.6HA部署-创新互联
当前地址：
http://cdkjz.cn/article/dejgpp.html
