參考
- Cloudera Enterprise 5.16.x
- Installing Cloudera Manager, CDH, and Managed Services
- Installation Path B - Manual Installation Using Cloudera Manager Packages
一. 環境
1.1 系統節點信息
Hostname | IP | CPU(cores) | Memory(GB) | OS | Service | Remark |
---|---|---|---|---|---|---|
master | 172.30.200.75 | 2 | 4 | centos 7.5 | jdk, cloudera-scm-server, mysql | 實際部署后,master節點也需要部署cloudera-scm-agent 服務,或采用4個slave節點 |
slave01 | 172.30.200.76 | 2 | 4 | centos 7.5 | jdk, cloudera-scm-agent | |
slave02 | 172.30.200.77 | 2 | 4 | centos 7.5 | jdk, cloudera-scm-agent | |
slave03 | 172.30.200.78 | 2 | 4 | centos 7.5 | jdk, cloudera-scm-agent |
1.2 軟件版本
相關軟件放置在/usr/local/src/
目錄。
Soft | Version | Download | Remark |
---|---|---|---|
CM(Cloudera Manager) | cloudera-manager-centos7-cm5.16.1_x86_64.tar.gz |
bin包,根據版本下載 | |
CDH parcel | CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel |
根據版本下載 | 軟件安裝包 |
CDH parcel.sha | CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel.sha1 |
根據版本下載 | 軟件包hash碼 |
CDH manifest.json | manifest.json |
根據版本下載 | 版本說明文件 |
JDK | jdk-8u181-linux-x64.tar.gz |
bin包,根據版本下載 | |
MySQL | mysql-5.7.24-el7-x86_64.tar.gz |
bin包,根據版本下載 | 存放Cloudera Manager配置文件 |
MySQL-connector-Java | mysql-connector-java-8.0.13.jar |
jar包,根據版本下載 | JDBC |
二. 准備工作
2.1 部分預配置
所有節點執行以下操作:
- 永久關閉防火牆(
firewalld
或iptables
); - 永久關閉
selinux
; - 設置
ntp
。
2.2 設置hosts
- 所有節點設置/etc/hosts
cat << EOF >> /etc/hosts # hadoop nodes 172.30.200.75 master 172.30.200.76 slave01 172.30.200.77 slave02 172.30.200.78 slave03 EOF
2.3 免密訪問設置
-
在所有節點生成秘鑰
ssh-keygen -t rsa
-
在master節點生成
authorized_keys
文件cat /root/.ssh/id_rsa.pub >> /root/.ssh/authorized_keys
-
在master節點執行,將集群每個節點的公鑰
id_rsa.pub
放入master節點的authorized_keys
文件中for i in {1..3}; do ssh root@slave0$i cat /root/.ssh/id_rsa.pub >> /root/.ssh/authorized_keys; done
-
在master節點執行,將master節點的
authorized_keys
文件放置到集群每個節點的/root/.ssh/
目錄,依然命名位authorized_keys
- 最終的效果是集群中所有節點兩兩相互免密訪問;
- 注意:首次登陸時有
公鑰檢查
,可通過在/etc/ssh/ssh_config
文件中設置StrictHostKeyChecking no
繞過,或使用-o
參數跳過。
for i in {1..3}; do scp /root/.ssh/authorized_keys root@slave0$i:/root/.ssh/authorized_keys; done
2.4 安裝JDK
所有節點都需要安裝JDK。
-
安裝JDK
cd /usr/local/src tar -zxvf jdk-8u181-linux-x64.tar.gz mkdir -p /usr/java mv jdk1.8.0_181/ /usr/java/ # 設置軟鏈接,方便升級替換 ln -s /usr/java/jdk1.8.0_181/ /usr/java/current
-
設置變量
cat << EOF >> /etc/profile # JDK export JAVA_HOME=/usr/java/current export JRE_HOME=/usr/java/current/jre export PATH=$PATH:/usr/java/current/bin export CLASSPATH=./:/usr/java/current/lib:/usr/java/current/jre/lib EOF
-
驗證
# 加載變量 source /etc/profile # 驗證 java -version
2.5 安裝MySQL
只有master節點需要安裝MySQL。
-
卸載系統自帶的相關數據庫
rpm -qa | grep mysql rpm -qa | grep mariadb # --nodeps:不檢查依賴 rpm -e --nodeps mariadb-libs-5.5.56-2.el7.x86_64
-
部署MySQL
# 采用bin包部署,解壓后直接使用 cd /usr/local/src tar -zxvf mysql-5.7.24-el7-x86_64.tar.gz -C /usr/local/ mv /usr/local/mysql-5.7.24-el7-x86_64/ /usr/local/mysql
-
賬號與權限
# 添加賬號 groupadd mysql useradd -g mysql -s /sbin/nologin mysql # 創建"data"目錄 mkdir /data # 賦權 chown -R mysql:mysql /usr/local/mysql/ chown -R mysql:mysql /data/
-
初始化MySQL
- 獲取
root@localhost
賬號的初始密碼Hqe6x<Re4jhK
cd /usr/local/mysql/ bin/mysqld --initialize --user=mysql --basedir=/usr/local/mysql --datadir=/data
- 獲取
-
設置變量
cat << EOF >> /etc/profile # MySQL export PATH=$PATH:/usr/local/mysql/bin EOF # 加載變量 source /etc/profile # 軟鏈接 ln -s /usr/local/mysql/lib/mysql /usr/lib/mysql ln -s /usr/local/mysql/include/mysql /usr/include/mysql
-
設置開機啟動
# 復制開機啟動腳本到系統服務 cp /usr/local/mysql/support-files/mysql.server /etc/rc.d/init.d/mysqld chown mysql:mysql /etc/rc.d/init.d/mysqld # 修改默認的"basedir"與"datadir" vim /etc/rc.d/init.d/mysqld basedir=/usr/local/mysql datadir=/data # 添加開機啟動腳本 chkconfig --add mysqld chkconfig --level 35 mysqld on
-
文件路徑:log && pid && socket
# 日志路徑 mkdir -p /var/log/mysqld touch /var/log/mysqld/mysqld.log chown -R mysql:mysql /var/log/mysqld/ # pid路徑 mkdir -p /var/run/mysqld chown -R mysql:mysql /var/run/mysqld/ # socker路徑 mkdir -p /var/lib/mysqld chown -R mysql:mysql /var/lib/mysqld/ ln -s /var/lib/mysqld/mysql.sock /tmp/mysql.sock
-
設置my.cnf文件
# 注意"log-error","pid-file"與"socket"的路徑 mkdir -p /usr/local/mysql/etc cat << EOF >> /usr/local/mysql/etc/my.cnf [mysqld] character-set-server=utf8 max_connections = 3000 log-error=/var/log/mysqld/mysqld.log pid-file=/var/run/mysqld/mysqld.pid socket=/var/lib/mysqld/mysql.sock sql_mode=NO_ENGINE_SUBSTITUTION,STRICT_TRANS_TABLES [mysql] default-character-set=utf8 EOF # 軟鏈接 ln -s /usr/local/mysql/etc/my.cnf /etc/my.cnf # 賦權 chown -R mysql:mysql /usr/local/mysql/etc/
-
啟動MySQL服務
# 啟動服務 service mysqld start # 驗證 service mysqld status
-
設置MySQL賬號密碼與登陸權限
# 使用初始化密碼登陸 mysql -uroot -p # 修改密碼,注意不能使用"$"等特殊符號 set password=password('cdh12#hadoop'); flush privileges; # 遠程登陸權限 grant all privileges on *.* to 'root'@'%' identified by 'cdh12#hadoop' with grant option; flush privileges; # 查看賬號 select user, host, authentication_string from mysql.user;
三. Cloudera Manager安裝
3.1 部署CM Server & Agent
-
在所有節點創建CM目錄
/opt/cloudera-manager
mkdir /opt/cloudera-manager
-
在master節點向其余節點分發
cloudera-manager-centos7-cm5.16.1_x86_64.tar.gz
for i in {1..3}; do scp /usr/local/src/cloudera-manager-centos7-cm5.16.1_x86_64.tar.gz root@slave0$i:/usr/local/src/ ; done
-
在所有節點解壓
cloudera-manager-centos7-cm5.16.1_x86_64.tar.gz
到/opt/cloudera-manager/
目錄tar -zxvf /usr/local/src/cloudera-manager-centos7-cm5.16.1_x86_64.tar.gz -C /opt/cloudera-manager/
-
在所有slave節點修改
/opt/cloudera-manager/cm-5.16.1/etc/cloudera-scm-agent/config.ini
的server_host
參數為master節點ip或主機名sed -i "s/server_host=localhost/server_host=master/g" /opt/cloudera-manager/cm-5.16.1/etc/cloudera-scm-agent/config.ini
3.2 賬號 && 權限
- 在所有節點創建
cloudera-scm
賬號,這是CM相關服務使用的默認賬號# 禁止使用"cloudera-scm"賬號登陸 useradd --system --home=/opt/cloudera-manager/cm-5.16.1/run/cloudera-scm-server/ --no-create-home --shell=/bin/false --comment "Cloudera SCM User" cloudera-scm
- 在所有節點為
/opt/cloudera-manager
目錄賦權chown -R cloudera-scm:cloudera-scm /opt/cloudera-manager
3.3 設置開機啟動
- 在master節點設置系統服務
# 設置使用"cloudera-scm-server"為系統啟動服務 cp /opt/cloudera-manager/cm-5.16.1/etc/init.d/cloudera-scm-server /etc/rc.d/init.d/ chown cloudera-scm:cloudera-scm /etc/rc.d/init.d/cloudera-scm-server # 修改"CMF_DEFAULTS=${CMF_DEFAULTS:-/etc/default}"的路徑 vim /etc/rc.d/init.d/cloudera-scm-server CMF_DEFAULTS=/opt/cloudera-manager/cm-5.16.1/etc/default # 添加系統啟動服務 chkconfig --add cloudera-scm-server chkconfig --level 35 cloudera-scm-server on checkconfig --list
- 在所有slave節點設置系統服務
# 設置使用"cloudera-scm-agent"為系統啟動服務 cp /opt/cloudera-manager/cm-5.16.1/etc/init.d/cloudera-scm-agent /etc/rc.d/init.d/ chown cloudera-scm:cloudera-scm /etc/rc.d/init.d/cloudera-scm-agent # 修改"CMF_DEFAULTS=${CMF_DEFAULTS:-/etc/default}"的路徑"-/etc/default" vim /etc/rc.d/init.d/cloudera-scm-agent CMF_DEFAULTS=${CMF_DEFAULTS:/opt/cloudera-manager/cm-5.16.1/etc/default} # 添加系統啟動服務 chkconfig --add cloudera-scm-agent chkconfig --level 35 cloudera-scm-agent on checkconfig --list
3.4 初始化數據庫
-
在所有節點設置MySQL驅動(JDBC);
-
注意:部署JDBC在任意節點,則后續"CDH安裝配置"階段
Reports Manager
被分配在任意節點都可以cp /usr/local/src/mysql-connector-java-8.0.13.jar /opt/cloudera-manager/cm-5.16.1/share/cmf/lib/ chown cloudera-scm:cloudera-scm /opt/cloudera-manager/cm-5.16.1/share/cmf/lib/mysql-connector-java-8.0.13.jar
-
在master節點重啟MySQL服務
service mysqld restart
-
在任意節點初始化CM
-
注意:Cloudera服務需要的相關
database
如下:- 表中給出的是CM相關服務配置文件中默認的
database
與user
,但不是必須使用; database
在數據庫中可直接創建,但CM初始化時如果沒有database
,則自動創建。
Service Database User Cloudera Manager Server scm scm Activity Monitor amon amon Reports Manager rman rman Hue hue hue Hive Metastore Server metastore metastore Sentry Server sentry sentry Cloudera Navigator Audit Server nav nav Cloudera Navigator Metadata Server navms navms Oozie oozie oozie # 格式:scm_prepare_database.sh [options] (postgresql|mysql|oracle) database username [password] # scm_prepare_database.sh:創建與配置CMS需要的數據庫腳本,默認在"/opt/cloudera-manager/cm-5.16.1/share/cmf/schema/"目錄; # postgresql|mysql|oracle:必選項,數據庫類型; # database:必選項,針對postgresql|mysql,創建SCM數據庫;針對oracle,填寫sid; # username:必選項,SCM數據庫的賬號; # password:選填項,SCM數據庫的賬號密碼,如果不指定,會提示輸入; # options: # -h:數據庫主機ip或hostname,默認是"localhost"; # -u:數據庫賬號,需要具備增刪改查的權限,默認是"root"; # -p:賬號密碼,默認無密碼; # --scm-host:SCM server主機名,默認是"localhost" /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % scm scm scm_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % amon amon amon_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % rman rman rman_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % hue hue hue_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % metastore metastore metastore_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % sentry sentry sentry_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % nav nav nav_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % navms navms navms_pass /opt/cloudera-manager/cm-5.16.1/share/cmf/schema/scm_prepare_database.sh mysql -h master -uroot -pcdh12#hadoop --scm-host % oozie oozie oozie_pass
返回如下信息,表示配置成功
[main] INFO com.cloudera.enterprise.dbutil.DbCommandExecutor - Successfully connected to database. All done, your SCM database is configured correctly!
- 表中給出的是CM相關服務配置文件中默認的
3.5 創建本地parcel源
-
在master節點制作本地parcel源
# 創建本地parcel源目錄 mkdir -p /opt/cloudera/parcel-repo # 將parcel相關安裝包放置到"/opt/cloudera/parcel-repo"目錄; # 說明:"/opt/cloudera/parcel-repo"目錄可放置多套parcel安裝包; # 將"CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel.sha1"重命名為"CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel.sha",否則會重新下載"CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel"安裝包 mv /usr/local/src/CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel /opt/cloudera/parcel-repo/ mv /usr/local/src/CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel.sha1 /opt/cloudera/parcel-repo/CDH-5.16.1-1.cdh5.16.1.p0.3-el7.parcel.sha mv /usr/local/src/manifest.json /opt/cloudera/parcel-repo/ # 賦權 chown -R cloudera-scm:cloudera-scm /opt/cloudera/
-
在所有salve節點創建軟件安裝目錄
mkdir -p /opt/cloudera/parcels # 賦權 chown -R cloudera-scm:cloudera-scm /opt/cloudera/
3.6 啟動CM服務
- 在master節點啟動
cloudera-scm-server
服務# "cloudera-scm-server"啟動需要連接數據庫,監聽端口啟動會延遲 service cloudera-scm-server restart service cloudera-scm-server status -l # 通過啟動后的狀態查看,腳本需要執行"pstree"命令,需要安裝依賴包 yum install psmisc -y
- 在所有salve節點啟動
cloudera-scm-agent
服務yum install psmisc -y service cloudera-scm-agent restart service cloudera-scm-agent status -l
四. 安裝CDH
4.1 CDH安裝配置
- 瀏覽器訪問CM:http://172.30.200.75:7180
- 默認賬號/密碼:
admin/admin
- 最終用戶許可條款與條件
- 部署版本,選擇
Cloudera Enterprise 試用版
- 版本與服務說明
cloudera-scm-agent
正常啟動后,可發現相應的節點,指定集群服務的安裝節點
- 選擇CDH版本
- 集群安裝,如果本地
parcel
源配置正確,則"下載"階段瞬間完成,其余階段視節點數與內部網絡情況決定
- 檢查主機正確性(需要一些時間),針對slave節點有兩個優化建議:
- 1.建議將
/proc/sys/vm/swappiness
設置為最大值10
。swappiness
值控制操作系統嘗試交換內存的積極;swappiness=0
:表示最大限度使用物理內存,之后才是swap空間;swappiness=100
:表示積極使用swap分區,並且把內存上的數據及時搬遷到swap空間;- 如果是混合服務器,不建議完全禁用swap,可以嘗試降低
swappiness
。 - 臨時調整:
sysctl vm.swappiness=10
- 永久調整:
cat << EOF >> /etc/sysctl.conf # Adjust swappiness value vm.swappiness=10 EOF
- 2.已啟用透明大頁面壓縮,可能會導致重大性能問題,建議禁用此設置。
- 臨時調整:
echo never > /sys/kernel/mm/transparent_hugepage/defrag echo never > /sys/kernel/mm/transparent_hugepage/enabled
- 永久調整:
cat << EOF >> /etc/rc.d/rc.local # Disable transparent_hugepage echo never > /sys/kernel/mm/transparent_hugepage/defrag echo never > /sys/kernel/mm/transparent_hugepage/enabled EOF # centos7.x系統,需要為"/etc/rc.d/rc.local"文件賦予執行權限 chmod +x /etc/rc.d/rc.local
- 臨時調整:
- 1.建議將
- 集群設置,選擇服務進行安裝,這里選擇
自定義服務
的HDFS
服務
- 集群設置,自定義角色分配,默認即可
- 集群設置,
Reports Manager
數據庫名稱rman
,用戶名rman
- 集群設置,審核更改,默認即可
- 集群設置,首次運行
- 集群設置,服務安裝完成
- Cloudera Manager 主頁
- Cloudera Manager HDFS 服務匯總