hadoop研究
包下载
http://archive.cloudera.com/cdh5/cdh/4/
http://apache.fayea.com/hadoop/common/hadoop-2.6.4/hadoop-2.6.4.tar.gz
http://mirrors.hust.edu.cn/apache/zookeeper/zookeeper-3.4.8/zookeeper-3.4.8.tar.gz
http://apache.opencas.org/hbase/1.2.0/hbase-1.2.0-bin.tar.gz
http://download.oracle.com/otn-pub/java/jdk/8u73-b02/jdk-8u73-linux-x64.tar.gz
环境
10.200.140.58 hadoop-308.99bill.com #物理机 datanode zookeeper regionserver
10.200.140.59 hadoop-309.99bill.com #物理机 datanode zookeeper regionserver
10.200.140.60 hadoop-310.99bill.com #物理机 datanode zookeeper regionserver
10.200.140.45 hadoop-311.99bill.com #虚拟机 master
10.200.140.46 hadoop-312.99bill.com #虚拟机 second hmaster
修改主机名，禁用ipv6
cat /etc/profile
export JAVA_HOME=/opt/jdk1.7.0_80/
PATH=$PATH:$JAVA_HOME/bin:$JAVA_HOME/jre/bin
CLASSPATH=.:$JAVA_HOME/lib:$JAVA_HOME/jre/lib
export JAVA_HOME
export PATH
export CLASSPATH
HADOOP_BASE=/opt/oracle/hadoop
HADOOP_HOME=/opt/oracle/hadoop
YARN_HOME=/opt/oracle/hadoop
PATH=$HADOOP_BASE/bin:$PATH
export HADOOP_BASE PATH
10.200.140.45 能够免密登陆
[oracle@hadoop-311 hadoop]$ cat core-site.xml
[oracle@hadoop-311 hadoop]$ cat hdfs-site.xml
[oracle@hadoop-311 hadoop]$ cat mapred-site.xml
定义 datanode
[oracle@hadoop-311 hadoop]$ cat slaves
hadoop-308.99bill.com
hadoop-309.99bill.com
hadoop-310.99bill.com
hadoop-env.sh
export HADOOP_LOG_DIR=$HADOOP_HOME/logs
export HADOOP_PID_DIR=/opt/oracle/hadoop
export HADOOP_SECURE_DN_PID_DIR=/opt/oracle/hadoop
export JAVA_HOME=/opt/jdk1.7.0_80/
export HADOOP_HEAPSIZE=6000
exec_time=`date +'%Y%m%d-%H%M%S'`
export HADOOP_NAMENODE_OPTS="-Xmx6g ${HADOOP_NAMENODE_OPTS}"
export HADOOP_SECONDARYNAMENODE_OPTS="-Xmx6g ${HADOOP_SECONDARYNAMENODE_OPTS}"
export HADOOP_DATANODE_OPTS="-server -Xmx6000m -Xms6000m -Xmn1000m -XX:PermSize=128M -XX:MaxPermSize=128M -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:$HADOOP_LOG_DIR/gc-$(hostname)-datanode-${exec_time}.log -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:CMSInitiatingOccupancyFraction=70 -XX:+UseCMSCompactAtFullCollection -XX:CMSFullGCsBeforeCompaction=10 -XX:+CMSClassUnloadingEnabled -XX:+CMSParallelRemarkEnabled -XX:+UseCMSInitiatingOccupancyOnly -XX:TargetSurvivorRatio=90 -XX:MaxTenuringThreshold=20"
[oracle@hadoop-311 hadoop]$ cat yarn-site.xml
启动hadoop集群
第一次执行，需要格式化namenode,以后启动不需要执行此步骤。
hadoop/bin/hadoop -format
然后启动hadoop
hadoop/sbin/start-all.sh
启动完成后，如果没有什么错误，执行jps查询一下当前进程，NameNode是Hadoop Master进程，SecondaryNameNode，ResourceManager是Hadoop进程。
[oracle@hadoop-311 hadoop]$ jps
13332 Jps
5430 NameNode
5719 ResourceManager
三、ZooKeeper集群安装
1. 解压缩zookeeper-3.4.8.tar.gz并重命名zookeeper, 进入zookeeper/conf目录,cp zoo_sample.cfg zoo.cfg 并编辑
[oracle@hadoop-308 conf]$ cat zoo.cfg
# The number of milliseconds of each tick
tickTime=2000
maxClientCnxns=0
# The number of ticks that the initial
# synchronization phase can take
initLimit=50
# The number of ticks that can pass between
# sending a request and getting an acknowledgement
syncLimit=5
# the directory where the snapshot is stored.
# 保留快照数
autopurge.snapRetainCount=2
# Purge task interval in hours
# 清理快照时间间隔（小时）
autopurge.purgeInterval=84
dataDir=/opt/hadoop/zookeeperdata
# the port at which the clients will connect
clientPort=2181
server.1=hadoop-308:2888:3888
server.2=hadoop-309:2888:3888
server.3=hadoop-310:2888:3888
2. 新建并编辑myid文件
1
mkdir /opt/hadoop/zookeeperdata
echo "1" > /opt/hadoop/zookeeperdata/myid
3. 然后同步zookeeper到其他两个节点,然后在其他节点需要修改myid为相应的数字。
启动 zookeeper
cd /opt/oracle/zookeeper
./bin/zkServer.sh start
[oracle@hadoop-308 tools]$ jps
11939 Jps
4373 DataNode
8579 HRegionServer
四、HBase集群的安装和配置
1. 解压缩hbase-1.2.0-bin.tar.gz并重命名为hbase, 编辑/hbase/conf/hbase-env.sh
export HBASE_MANAGES_ZK=false
export HBASE_HEAPSIZE=4000
export JAVA_HOME=/opt/jdk1.7.0_80/
[oracle@hadoop-311 conf]$ cat hbase-site.xml
Zookeeper session expired will force regionserver exit.
Enable this will make the regionserver restart.
分发hbase到其他4个节点
五、启动集群
1. 启动zookeeper
1
zookeeper/bin/zkServer.sh start
2. 启动Hadoop
$ hadoop/sbin/start-all.sh
修改hbase/conf/hbase-site.xml
[oracle@hadoop-311 conf]$ cat hbase-site.xml
Zookeeper session expired will force regionserver exit.
Enable this will make the regionserver restart.
hbase-env.sh
export JAVA_HOME=/opt/jdk1.7.0_80/
export HBASE_CLASSPATH=/opt/oracle/hadoop/conf
export HBASE_HEAPSIZE=4000
export HBASE_OPTS="-XX:PermSize=512M -XX:MaxPermSize=512M -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:CMSInitiatingOccupancyFraction=70 -XX:+UseCMSCompactAtFullCollection -XX:CMSFullGCsBeforeCompaction=10 -XX:+CMSClassUnloadingEnabled -XX:+CMSParallelRemarkEnabled -XX:+UseCMSInitiatingOccupancyOnly -XX:TargetSurvivorRatio=90 -XX:MaxTenuringThreshold=20"
exec_time=`date +'%Y%m%d-%H%M%S'`
export HBASE_MASTER_OPTS="-Xmx4096m -Xms4096m -Xmn128m -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:$HBASE_HOME/logs/gc-$(hostname)-master-${exec_time}.log"
export HBASE_REGIONSERVER_OPTS="-Xmx8192m -Xms8192m -Xmn512m -verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:$HBASE_HOME/logs/gc-$(hostname)-regionserver-${exec_time}.log"
export HBASE_MANAGES_ZK=fals
[oracle@hadoop-311 conf]$ cat regionservers
hadoop-308
hadoop-309
hadoop-310
分发到其他四台
cd /opt/oracle/hbase
sh bin/start-hbase.sh
[oracle@hadoop-311 bin]$ ./hbase shell
16/03/23 20:20:47 WARN conf.Configuration: hadoop.native.lib is deprecated. Instead, use io.native.lib.available
HBase Shell; enter 'help
Type "exit
Version 0.94.15-cdh5.7.1, r, Tue Nov 18 08:42:59 PST 2014
hbase(main):001:0> status
SLF4J: Class path contains multiple SLF4J bindings.
SLF4J: Found binding in [jar:file:/opt/oracle/hbase/lib/slf4j-log4j12-1.6.1.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: Found binding in [jar:file:/opt/oracle/hadoop/share/hadoop/common/lib/slf4j-log4j12-1.6.1.jar!/org/slf4j/impl/StaticLoggerBinder.class]
SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.
16/03/23 20:20:52 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
3 servers, 0 dead, 0.6667 average load
10. 常见问题
10.1. Namenode非正常关闭
在所有的hadoop环境机器上用jps命令，把所有的进程列出，然后kill掉，再按照启动顺序启动
10.2. Datanode非正常关闭
在namenode上启动HDFS
运行hadoop/bin/start-all.sh
如果Datanode同时是zookeeper，还需要启动zookeeper
在该datanode上运行zookeeper/bin/zkServer.sh start。
在namenode上启动Hbase
运行hbase/bin/start-hbase.sh
http://10.200.140.46:60010/master-status
10.3. 停止一台非master的服务器
在该台服务器上运行：
hadoop/bin/hadoop-daemon.sh stop datanode
hadoop/bin/hadoop-daemon.sh stop tasktracker
hbase/bin/hbase-daemon.sh stop regionserver
在http://10.200.140.45:50070/dfshealth.jsp查看该节点是否已经变成dead nodes，变成dead nodes之后，就可以停止该台服务器
在刚停止服务的时候，看到的截图如下：
当停止服务成功，看到的截图如下：
重启服务器以后，在hadoop001上运行，启动服务：
hadoop/bin/start-all.sh
hbase/bin/start-hbase.sh
11. 监控端口
11.1. Namenode监控端口(hadoop001)：
60010,60000,50070,50030,9000,9001,10000
11.2. zookeeper监控端口(hadoop003,hadoop004,hadoop005)
2181
11.3. Datanode监控端口(hadoop003,hadoop004,hadoop005,hadoop006,hadoop007)
60030,50075
12、HDFS 上传文件不均衡和Balancer太慢的问题
Hmaster 有个start-balancer.sh
###########迁移方案
先在新机房准备一套新的hadoop环境
###hadoop迁移-hbase
1 确定新hbase可以正常运行，并且两个集群之间的机器都可以用机器名互相访问到 ok
2 停掉新hbase ok
3 在两个集群任何hadoop机器运行下面的命令
./hadoop distcp -bandwidth 10 -m 3 hdfs://hadoop001.99bill.com:9000/hbase/if_fss_files hdfs://hadoop-312.99bill.com:9000/hbase/if_fss_files
4 使用附件的脚本，运行
hbase org.jruby.Main ~/add_table.rb /hbase/if_fss_files
5 启动新hbase
###hadoop迁移-hadoop数据迁移
########整理hadoop文件,对于打包失败的重新打包
如2014-07-24执行
./hdfs dfs -rm -r /fss/2014-07-24
./hdfs dfs -rm -r /fss/2014-07-24.har
./hdfs dfs -mv /fss/2014-07-24a.har /fss/2014-07-24.har
##从远程fss系统同步到新机房本地
./hdfs dfs -copyToLocal hdfs://hadoop001.99bill.com:9000/fss/2015-04-08.har /opt/sdb/hadoop/tmp/
####从新机房本地导入fss系统
./hdfs dfs -copyFromLocal /opt/sdb/hadoop/tmp/2015-04-08.har /fss/
sleep 5
./hdfs dfs -copyFromLocal /opt/sdb/hadoop/tmp/2015-06/03-30.har /fss/2015-06
