Hadoop 搭建高可用完全分布式集群

部署规划

主机	用途	IP
rm01.hadoop.com	ResourceManager01	192.168.137.11
nn01.hadoop.com	NameNode01 、 DFSZKFailoverController	192.168.137.12
rm02.hadoop.com (backup resourcemanager)	ResourceManager02	192.168.137.13
nn02.hadoop.com (backup namenode)	NameNode02 、 DFSZKFailoverController	192.168.137.14
dn01.hadoop.com	Dat a Node 、 NodeManager 、 QuorumPeerMain 、 JournalNode	192.168.137.21
dn02.hadoop.com	Dat a Node 、 NodeManager 、 QuorumPeerMain 、 JournalNode	192.168.137.22
dn03.hadoop.com	Dat a Node 、 NodeManager 、 QuorumPeerMain 、 JournalNode	192.168.137.23

[hadoop@dn01 ~]$ tar -zxf /nfs_share/software/zookeeper-3.4.11.tar.gz -C ~

[hadoop@dn01 ~]$ vi .bashrc

export ZOOKEEPER_HOME=/home/hadoop/zookeeper-3.4.11

export PATH=$PATH:/home/hadoop/zookeeper-3.4.11/bin

[hadoop@dn01 ~]$ source .bashrc

[hadoop@dn01 ~]$ cd zookeeper-3.4.11/conf

[hadoop@dn01 conf]$ mv zoo_sample.cfg zoo.cfg

[hadoop@dn01 conf]$ vi zoo.cfg

dataLogDir=/home/hadoop/zookeeper-3.4.11/log

dataDir=/home/hadoop/zookeeper-3.4.11/data

server.1=192.168.137.21:2888:3888

server.2=192.168.137.22:2888:3888

server.3=192.168.137.23:2888:3888

[hadoop@dn01 conf]$ cd ..

[hadoop@dn01 zookeeper-3.4.11]$ mkdir data && mkdir log && cd data && echo "1">>myid

[hadoop@dn01 zookeeper-3.4.11]$ cd

[hadoop@dn01 ~]$ scp -r zookeeper-3.4.11 dn02.hadoop.com:/home/hadoop

[hadoop@dn01 ~]$ scp -r zookeeper-3.4.11 dn03.hadoop.com:/home/hadoop

[hadoop@dn01 ~]$ ssh [email protected] 'cd /home/hadoop/zookeeper-3.4.11/data && echo "2">myid'

[hadoop@dn01 ~]$ ssh [email protected] 'cd /home/hadoop/zookeeper-3.4.11/data && echo "3">myid'

[hadoop@dn01 ~]$ zkServer.sh start

[hadoop@dn02 ~]$ zkServer.sh start

[hadoop@dn03 ~]$ zkServer.sh start

[hadoop@dn01 ~]$ zkServer.sh status

[hadoop@dn02 ~]$ zkServer.sh status

[hadoop@dn03 ~]$ zkServer.sh status

[hadoop@dn01 ~]$ cd hadoop-2.9.0 && mkdir journal

[hadoop@dn02 ~]$ cd hadoop-2.9.0 && mkdir journal

[hadoop@dn03 ~]$ cd hadoop-2.9.0 && mkdir journal

[hadoop@nn01 ~]$ cd hadoop-2.9.0/etc/hadoop/

[hadoop@nn01 hadoop]$ vi core-site.xml

fs.defaultFS         hdfs://ns1                   hadoop.tmp.dir         file:/home/hadoop/hadoop-2.9.0/tmp                   io.file.buffer.size         131072                      hadoop.proxyuser.hadoop.hosts            *                      hadoop.proxyuser.hadoop.groups         *                        ha.zookeeper.quorum         dn01.hadoop.com:2181,dn02.hadoop.com:2181,dn03.hadoop.com:2181                   ha.zookeeper.session-timeout.ms         1000

[hadoop@nn01 hadoop]$ vi hdfs-site.xml

dfs.namenode.name.dir         /home/hadoop/hadoop-2.9.0/dfs/name                   dfs.datanode.data.dir         /home/hadoop/hadoop-2.9.0/dfs/data                   dfs.blocksize         64m                   dfs.replication         3                   dfs.webhdfs.enabled         true                   dfs.permissions          false                   dfs.permissions.enabled          false                        dfs.nameservices          ns1                        dfs.ha.namenodes.ns1          nn1,nn2                   dfs.namenode.rpc-address.ns1.nn1           nn01.hadoop.com:8020                       dfs.namenode.rpc-address.ns1.nn2           nn02.hadoop.com:8020                       dfs.namenode.servicerpc-address.ns1.nn1           nn01.hadoop.com:53310                       dfs.namenode.servicerpc-address.ns1.nn2           nn02.hadoop.com:53310                       dfs.namenode.http-address.ns1.nn1           nn01.hadoop.com:50070                       dfs.namenode.http-address.ns1.nn2           nn02.hadoop.com:50070                   dfs.namenode.shared.edits.dir.ns1.nn1         qjournal:// dn01.hadoop.com:8485;dn02.hadoop.com:8485;dn03.hadoop.com:8485/ns1                   dfs.namenode.shared.edits.dir.ns1.nn2         qjournal:// dn01.hadoop.com:8485;dn02.hadoop.com:8485;dn03.hadoop.com:8485/ns1                   dfs.ha.automatic-failover.enabled.ns1         true                   dfs.client.failover.proxy.provider.ns1         org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider                   dfs.ha.fencing.methods         sshfence                   dfs.ha.fencing.ssh.private-key-files         /home/hadoop/.ssh/id_rsa                   dfs.ha.fencing.ssh.connect-timeout         30000                   dfs.journalnode.edits.dir         /home/hadoop/hadoop-2.9.0/journal                   ha.failover-controller.cli-check.rpc-timeout.ms         50000                   ipc.client.connect.timeout         60000                   dfs.image.transfer.bandwidthPerSec         4194304

[hadoop@nn01 hadoop]$ vi   mapred -site.xml

mapreduce.framework.name         yarn                   mapreduce.jobhistory.address         rm01.hadoop.com:10020                   mapreduce.jobhistory.webapp.address         rm01.hadoop.com:19888

[hadoop@nn01 hadoop]$ vi   yarn-site.xml

yarn.nodemanager.aux-services         mapreduce_shuffle                  yarn.nodemanager.aux-services.mapreduce.shuffle.class        org.apache.hadoop.mapred.ShuffleHandler                   yarn.resourcemanager.connect.retry-interval.ms         2000                   yarn.resourcemanager.scheduler.class         org.apache.hadoop.yarn.server.resourcemanager.scheduler.fair.FairScheduler                   yarn.app.mapreduce.am.scheduler.connection.wait.interval-ms         5000                     yarn.resourcemanager.recovery.enabled     true           yarn.resourcemanager.store.class     org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore           yarn.resourcemanager.zk-address     dn01.hadoop.com:2181,dn02.hadoop.com:2181,dn03.hadoop.com:2181           yarn.resourcemanager.zk-state-store.parent-path     /rmstore           yarn.resourcemanager.zk-num-retries     500           yarn.resourcemanager.zk-retry-interval-ms     2000           yarn.resourcemanager.zk-timeout-ms     10000           yarn.resourcemanager.zk-acl     world:anyone:rwcda           yarn.resourcemanager.am.max-attempts     2                 yarn.resourcemanager.cluster-id     yarn-cluster           yarn.resourcemanager.ha.enabled     true           yarn.resourcemanager.ha.rm-ids     rm1,rm2           yarn.resourcemanager.hostname.rm1     rm01.hadoop.com           yarn.resourcemanager.hostname.rm2     rm02.hadoop.com           yarn.resourcemanager.ha.automatic-failover.enabled     true           yarn.resourcemanager.ha.automatic-failover.embedded     true                 yarn.resourcemanager.address.rm1     rm01.hadoop.com:8032           yarn.resourcemanager.scheduler.address.rm1     rm01.hadoop.com:8030           yarn.resourcemanager.admin.address.rm1     rm01.hadoop.com:8033           yarn.resourcemanager.resource-tracker.address.rm1     rm01.hadoop.com:8031           yarn.resourcemanager.webapp.address.rm1     rm01.hadoop.com:8088           yarn.resourcemanager.webapp.https.address.rm1     rm01.hadoop.com:8090                 yarn.resourcemanager.address.rm2     rm02.hadoop.com:8032           yarn.resourcemanager.scheduler.address.rm2     rm02.hadoop.com:8030           yarn.resourcemanager.admin.address.rm2     rm02.hadoop.com:8033           yarn.resourcemanager.resource-tracker.address.rm2     rm02.hadoop.com:8031           yarn.resourcemanager.webapp.address.rm2     rm02.hadoop.com:8088           yarn.resourcemanager.webapp.https.address.rm2     rm02.hadoop.com:8090

[hadoop@nn01 hadoop]$ vi   slaves

dn01.hadoop.com dn02.hadoop.com dn03.hadoop.com

[hadoop@nn01 ~]$ hdfs zkfc -formatZK

启动journalnode节点用于namenode主备数据同步

[hadoop@dn01 ~]$ hadoop-daemon.sh start journalnode

[hadoop@dn02 ~]$ hadoop-daemon.sh start journalnode

[hadoop@dn03 ~]$ hadoop-daemon.sh start journalnode

启动主namenode

[hadoop@nn01 ~]$ hdfs namenode -format -clusterId c1

[hadoop@nn01 ~]$ hadoop-daemon.sh start namenode

启动备用namenode

[hadoop@nn02 ~]$ hdfs namenode -bootstrapStandby

[hadoop@nn02 ~]$ hadoop-daemon.sh start namenode

启动namenode故障转移程序

[hadoop@nn01 ~]$ hadoop-daemon.sh start zkfc

[hadoop@nn02 ~]$ hadoop-daemon.sh start zkfc

启动datanode

[hadoop@dn01 ~]$ hadoop-daemon.sh start datanode

[hadoop@dn02 ~]$ hadoop-daemon.sh start datanode

[hadoop@dn03 ~]$ hadoop-daemon.sh start datanode

启动主resoucemanager

[hadoop@rm01 ~]$ start-yarn.sh

启动备用resoucemanager

[hadoop@rm02 ~]$ yarn-daemon.sh start resourcemanager

http://nn01.hadoop.com:50070/dfshealth.html#tab-overview

http://nn02.hadoop.com:50070/dfshealth.html#tab-overview

http://rm01.hadoop.com:8088/cluster/cluster

http://rm02.hadoop.com:8088/cluster/cluster

HDFS HA 检验实验

[hadoop@nn01 ~]$ jps

2352 DFSZKFailoverController

2188 NameNode

3105 Jps

执行命令

[hadoop@nn01 ~]$ kill -9 2188

刷新页面，看到

说明切换成功。

ResourceManager HA 检验实验

[hadoop@rm01 ~]$ jps

1599 ResourceManager

1927 Jps

启动wordcount程序

kill掉主ResourceManager进程

[hadoop@rm01 ~]$ kill -9 1599

看控制台输出，可以看到备的ResourceManager被启用

说明切换成功。