Hadoop前面安装的集群是2.6版本，现在升级到2.7版本。
注意，这个集群上有运行Hbase，所以，升级前后，需要启停Hbase。
更多安装步骤，请参考：
升级步骤如下：
集群IP列表
Namenode: 192.168.143.46 192.168.143.103 Journalnode: 192.168.143.101 192.168.143.102 192.168.143.103 Datanode&Hbase regionserver: 192.168.143.196 192.168.143.231 192.168.143.182 192.168.143.235 192.168.143.41 192.168.143.127 Hbase master: 192.168.143.103 192.168.143.101 Zookeeper: 192.168.143.101 192.168.143.102 192.168.143.103
1. 首先确定hadoop运行的路径，将新版本的软件分发到每个节点的这个路径下，并解压。
# ll /usr/local/hadoop/ total 493244 drwxrwxr-x 9 root root 4096 Mar 21 2017 hadoop-release ->hadoop-2.6.0-EDH-0u1-SNAPSHOT-HA-SECURITY drwxr-xr-x 9 root root 4096 Oct 11 11:06 hadoop-2.7.1 -rw-r--r-- 1 root root 194690531 Oct 9 10:55 hadoop-2.7.1.tar.gz drwxrwxr-x 7 root root 4096 May 21 2016 hbase-1.1.3 -rw-r--r-- 1 root root 128975247 Apr 10 2017 hbase-1.1.3.tar.gz lrwxrwxrwx 1 root root 29 Apr 10 2017 hbase-release -> /usr/local/hadoop/hbase-1.1.3
由于是升级，配置文件完全不变，将原hadoop-2.6.0下的etc/hadoop路径完全拷贝/替换到hadoop-2.7.1下。
至此，升级前的准备就已经完成了。
下面开始升级操作过程。全程都是在一个中转机上执行的命令，通过shell脚本执行，省去频繁ssh登陆的操作。
## 停止hbase，hbase用户执行
2. 停止Hbase master，hbase用户执行
状态检查，确认master，先停standby master
http://192.168.143.101:16010/master-status
master: ssh -t -q 192.168.143.103 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ master" ssh -t -q 192.168.143.103 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.101 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ master" ssh -t -q 192.168.143.101 sudo su -l hbase -c "jps"
3. 停止Hbase regionserver，hbase用户执行
ssh -t -q 192.168.143.196 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver" ssh -t -q 192.168.143.231 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver" ssh -t -q 192.168.143.182 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver" ssh -t -q 192.168.143.235 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver" ssh -t -q 192.168.143.41 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver" ssh -t -q 192.168.143.127 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ stop\ regionserver"
检查运行状态
ssh -t -q 192.168.143.196 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.231 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.182 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.235 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.41 sudo su -l hbase -c "jps" ssh -t -q 192.168.143.127 sudo su -l hbase -c "jps"
## 停止服务--HDFS
4. 先确认，active的namenode,网页确认.后续要先启动这个namenode
https://192.168.143.46:50470/dfshealth.html#tab-overview
5. 停止NameNode，hdfs用户执行
NN: 先停standby namenode
ssh -t -q 192.168.143.103 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ namenode" ssh -t -q 192.168.143.46 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ namenode" 检查状态 ssh -t -q 192.168.143.103 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.46 sudo su -l hdfs -c "jps"
6. 停止DataNode，hdfs用户执行
ssh -t -q 192.168.143.196 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode" ssh -t -q 192.168.143.231 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode" ssh -t -q 192.168.143.182 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode" ssh -t -q 192.168.143.235 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode" ssh -t -q 192.168.143.41 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode" ssh -t -q 192.168.143.127 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ datanode"
7. 停止ZKFC，hdfs用户执行
ssh -t -q 192.168.143.46 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ zkfc" ssh -t -q 192.168.143.103 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ zkfc"
8.停止JournalNode，hdfs用户执行
JN: ssh -t -q 192.168.143.101 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ journalnode" ssh -t -q 192.168.143.102 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ journalnode" ssh -t -q 192.168.143.103 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ stop\ journalnode"
### 备份NameNode的数据,由于生产环境，原有的数据需要备份。以备升级失败回滚。
9. 备份namenode1
ssh -t -q 192.168.143.46 "cp -r /data1/dfs/name /data1/dfs/name.bak.20171011-2;ls -al /data1/dfs/;du -sm /data1/dfs/*" ssh -t -q 192.168.143.46 "cp -r /data2/dfs/name /data2/dfs/name.bak.20171011-2;ls -al /data1/dfs/;du -sm /data1/dfs/*"
10. 备份namenode2
ssh -t -q 192.168.143.103 "cp -r /data1/dfs/name /data1/dfs/name.bak.20171011-2;ls -al /data1/dfs/;du -sm /data1/dfs/*"
11. 备份journal
ssh -t -q 192.168.143.101 "cp -r /data1/journalnode /data1/journalnode.bak.20171011;ls -al /data1/dfs/;du -sm /data1/*" ssh -t -q 192.168.143.102 "cp -r /data1/journalnode /data1/journalnode.bak.20171011;ls -al /data1/dfs/;du -sm /data1/*" ssh -t -q 192.168.143.103 "cp -r /data1/journalnode /data1/journalnode.bak.20171011;ls -al /data1/dfs/;du -sm /data1/*"
journal路径，可以查看hdfs-site.xml文件
dfs.journalnode.edits.dir: /data1/journalnode
### 升级相关
12. copy文件（已提前处理，参考第一步）
切换软连接到2.7.1版本
ssh -t -q $h "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release"
13. 切换文件软链接,root用户执行
ssh -t -q 192.168.143.46 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.103 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.101 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.102 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.196 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.231 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.182 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.235 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.41 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release" ssh -t -q 192.168.143.127 "cd /usr/local/hadoop; rm hadoop-release; ln -s hadoop-2.7.1 hadoop-release"
确认状态
ssh -t -q 192.168.143.46 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.103 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.101 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.102 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.196 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.231 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.182 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.235 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.41 "cd /usr/local/hadoop; ls -al" ssh -t -q 192.168.143.127 "cd /usr/local/hadoop; ls -al"
### 启动HDFS，hdfs用户执行
14. 启动JournalNode
JN: ssh -t -q 192.168.143.101 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ journalnode" ssh -t -q 192.168.143.102 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ journalnode" ssh -t -q 192.168.143.103 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ journalnode"
ssh -t -q 192.168.143.101 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.102 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.103 sudo su -l hdfs -c "jps"
15. 启动第一个NameNode
ssh 192.168.143.46 su - hdfs /usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh start namenode -upgrade
16. 确认状态,在状态完全OK之后，才可以启动另一个namenode
https://192.168.143.46:50470/dfshealth.html#tab-overview
17. 启动第一个ZKFC
su - hdfs /usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh start zkfc 192.168.143.46
18. 启动第二个NameNode
ssh 192.168.143.103 su - hdfs /usr/local/hadoop/hadoop-release/bin/hdfs namenode -bootstrapStandby /usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh start namenode
19. 启动第二个ZKFC
ssh 192.168.143.103 su - hdfs /usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh start zkfc
20. 启动DataNode
ssh -t -q 192.168.143.196 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode" ssh -t -q 192.168.143.231 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode" ssh -t -q 192.168.143.182 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode" ssh -t -q 192.168.143.235 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode" ssh -t -q 192.168.143.41 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode" ssh -t -q 192.168.143.127 sudo su -l hdfs -c "/usr/local/hadoop/hadoop-release/sbin/hadoop-daemon.sh\ start\ datanode"
确认状态
ssh -t -q 192.168.143.196 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.231 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.182 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.235 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.41 sudo su -l hdfs -c "jps" ssh -t -q 192.168.143.127 sudo su -l hdfs -c "jps"
21. 一切正常之后，启动hbase, hbase用户执行
启动hbase master，最好先启动原来的active master。
ssh -t -q 192.168.143.101 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ master" ssh -t -q 192.168.143.103 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ master"
启动Hbase regionserver
ssh -t -q 192.168.143.196 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver" ssh -t -q 192.168.143.231 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver" ssh -t -q 192.168.143.182 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver" ssh -t -q 192.168.143.235 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver" ssh -t -q 192.168.143.41 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver" ssh -t -q 192.168.143.127 sudo su -l hbase -c "/usr/local/hadoop/hbase-release/bin/hbase-daemon.sh\ start\ regionserver"
22. Hbase region需要手动Balance开启、关闭
需要登录HBase Shell运行如下命令
开启
balance_switch true
关闭
balance_switch false
23. 本次不执行，系统运行一周，确保系统运行稳定，再执行Final。
注意：这期间，磁盘空间可能会快速增长。在执行完final之后，会释放一部分空间。
Finallize upgrade: hdfs dfsadmin -finalizeUpgrade
本文出自 “小小狗窝” 博客，请务必保留此出处http://hsbxxl.blog.51cto.com/181620/1976472
原文：http://hsbxxl.blog.51cto.com/181620/1976472