Ansible实战：部署分布式日志系统

本节内容：

• 背景
• 分布式日志系统架构图
• 创建和使用roles
  • JDK 7 role
  • JDK 8 role
  • Zookeeper role
  • Kafka role
  • Elasticsearch role
  • MySQL role
  • Nginx role
  • Redis role
  • Hadoop role
  • Spark role

一、背景

产品组在开发一个分布式日志系统，用的组件较多，单独手工部署一各个个软件比较繁琐，花的时间比较长，于是就想到了使用ansible playbook + roles进行部署，效率大大提高。

二、分布式日志系统架构图

三、创建和使用roles

每一个软件或集群都创建一个单独的角色。

[root@node1 ~]# mkdir -pv ansible_playbooks/roles/{db_server,web_server,redis_server,zk_server,kafka_server,es_server,tomcat_server,flume_agent,hadoop,spark,hbase,hive,jdk7,jdk8}/{tasks,files,templates,meta,handlers,vars}

3.1. JDK7 role

[root@node1 jdk7]# pwd
/root/ansible_playbooks/roles/jdk7
[root@node1 jdk7]# ls
files  handlers  meta  tasks  templates  vars

1. 上传软件包

将jdk-7u80-linux-x64.gz上传到files目录下。

2. 编写tasks

[root@node1 jdk7]# vim tasks/main.yml 
- name: mkdir necessary catalog                                                                                                               
  file: path=/usr/java state=directory mode=0755
- name: copy and unzip jdk 
  unarchive: src={{jdk_package_name}} dest=/usr/java/
- name: set env 
  lineinfile: dest={{env_file}} insertafter="{{item.position}}" line="{{item.value}}" state=present
  with_items:
  - {position: EOF, value: "
"}
  - {position: EOF, value: "export JAVA_HOME=/usr/java/{{jdk_version}}"}
  - {position: EOF, value: "export PATH=$JAVA_HOME/bin:$PATH"}
  - {position: EOF, value: "export CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar"}
- name: enforce env 
  shell: source {{env_file}}

3. 编写vars

[root@node1 jdk7]# vim vars/main.yml 
jdk_package_name: jdk-7u80-linux-x64.gz                                                                                                       
env_file: /etc/profile
jdk_version: jdk1.7.0_80

4. 使用角色 

在roles同级目录，创建一个jdk.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim jdk.yml 
- hosts: jdk
  remote_user: root
  roles:
  - jdk7

运行playbook安装JDK7：

[root@node1 ansible_playbooks]# ansible-playbook jdk.yml 

使用jdk7 role可以需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.2 JDK8 role

[root@node1 jdk8]# pwd
/root/ansible_playbooks/roles/jdk8
[root@node1 jdk8]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将jdk-8u73-linux-x64.gz上传到files目录下。 

2. 编写tasks

[root@node1 jdk8]# vim tasks/main.yml 
- name: mkdir necessary catalog                                                                                                               
  file: path=/usr/java state=directory mode=0755
- name: copy and unzip jdk 
  unarchive: src={{jdk_package_name}} dest=/usr/java/
- name: set env 
  lineinfile: dest={{env_file}} insertafter="{{item.position}}" line="{{item.value}}" state=present
  with_items:
  - {position: EOF, value: "
"}
  - {position: EOF, value: "export JAVA_HOME=/usr/java/{{jdk_version}}"}
  - {position: EOF, value: "export PATH=$JAVA_HOME/bin:$PATH"}
  - {position: EOF, value: "export CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar"}
- name: enforce env 
  shell: source {{env_file}}

3. 编写vars

[root@node1 jdk8]# vim vars/main.yml 
jdk_package_name: jdk-8u73-linux-x64.gz                                                                                                       
env_file: /etc/profile
jdk_version: jdk1.8.0_73

4. 使用角色

在roles同级目录，创建一个jdk.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim jdk.yml 
- hosts: jdk
  remote_user: root
  roles:
  - jdk8

运行playbook安装JDK8：

[root@node1 ansible_playbooks]# ansible-playbook jdk.yml 

使用jdk8 role可以需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.3 Zookeeper role

Zookeeper集群节点配置好/etc/hosts文件，配置集群各节点主机名和ip地址的对应关系。

[root@node1 zk_server]# pwd
/root/ansible_playbooks/roles/zk_server
[root@node1 zk_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将zookeeper-3.4.6.tar.gz和clean_zklog.sh上传到files目录。clean_zklog.sh是清理Zookeeper日志的脚本。

2. 编写tasks

[root@node1 zk_server]# vim tasks/main.yml 
- name: install zookeeper                                                                                                                     
  unarchive: src=zookeeper-3.4.6.tar.gz dest=/usr/local/
- name: install configuration file for zookeeper
  template: src=zoo.cfg.j2 dest=/usr/local/zookeeper-3.4.6/conf/zoo.cfg
- name: add myid file
  shell: echo {{ myid }} > /usr/local/zookeeper-3.4.6/dataDir/myid
- name: copy script to clear zookeeper logs.
  copy: src=clean_zklog.sh dest=/usr/local/zookeeper-3.4.6/clean_zklog.sh mode=755
- name: crontab task
  cron: name="clear zk logs" weekday="0" hour="0" minute="0" job="/usr/local/zookeeper-3.4.6/clean_zklog.sh"
- name: start zookeeper
  shell: /usr/local/zookeeper-3.4.6/bin/zkServer.sh start
  tags:
  - start

3. 编写templates

将zookeeper-3.4.6.tar.gz包中的默认配置文件上传到../roles/zk_server/templates/目录下，重命名为zoo.cfg.j2，并修改其中的内容。

[root@node1 ansible_playbooks]# vim roles/zk_server/templates/zoo.cfg.j2

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不在解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 zk_server]# vim vars/main.yml 
server1_hostname: hadoop27                                                                                                                    
server2_hostname: hadoop28
server3_hostname: hadoop29

另外在tasks中还使用了个变量{{myid}}，该变量每台主机的值是不一样的，所以定义在了/etc/ansible/hosts文件中:

[zk_servers]
172.16.206.27 myid=1
172.16.206.28 myid=2
172.16.206.29 myid=3

5. 设置主机组

/etc/ansible/hosts文件：

[zk_servers]
172.16.206.27 myid=1
172.16.206.28 myid=2
172.16.206.29 myid=3

6. 使用角色

在roles同级目录，创建一个zk.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim zk.yml 
- hosts: zk_servers
  remote_user: root
  roles:
  - zk_server

运行playbook安装Zookeeper集群：

[root@node1 ansible_playbooks]# ansible-playbook zk.yml 

使用zk_server role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.4 Kafka role

[root@node1 kafka_server]# pwd
/root/ansible_playbooks/roles/kafka_server
[root@node1 kafka_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将kafka_2.11-0.9.0.1.tar.gz、kafka-manager-1.3.0.6.zip和clean_kafkalog.sh上传到files目录。clean_kafkalog.sh是清理kafka日志的脚本。

2. 编写tasks

[root@node1 kafka_server]# vim tasks/main.yml 
- name: copy and unzip kafka
  unarchive: src=kafka_2.11-0.9.0.1.tgz dest=/usr/local/
- name: install configuration file for kafka
  template: src=server.properties.j2 dest=/usr/local/kafka_2.11-0.9.0.1/config/server.properties
- name: copy script to clear kafka logs.
  copy: src=clean_kafkalog.sh dest=/usr/local/kafka_2.11-0.9.0.1/clean_kafkalog.sh mode=755                                                   
- name: crontab task                                                         
  cron: name="clear kafka logs" weekday="0" hour="0" minute="0" job="/usr/local/kafka_2.11-0.9.0.1/clean_kafkalog.sh"
- name: start kafka                            
  shell: JMX_PORT=9997 /usr/local/kafka_2.11-0.9.0.1/bin/kafka-server-start.sh -daemon /usr/local/kafka_2.11-0.9.0.1/config/server.properties &                                              
  tags:                                        
  - start                                      
- name: copy and unizp kafka-manager           
  unarchive: src=kafka-manager-1.3.0.6.zip dest=/usr/local/
  when: ansible_default_ipv4['address'] == "{{kafka_manager_ip}}"
- name: install configuration file for kafka-manager
  template: src=application.conf.j2 dest=/usr/local/kafka-manager-1.3.0.6/conf/application.conf
  when: ansible_default_ipv4['address'] == "{{kafka_manager_ip}}"
- name: start kafka-manager                    
  shell: nohup /usr/local/kafka-manager-1.3.0.6/bin/kafka-manager &
  when: ansible_default_ipv4['address'] == "{{kafka_manager_ip}}"
  tags:                                        
  - kafkaManagerStart

3. 编写templates

[root@node1 kafka_server]# vim templates/server.properties.j2 

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不再解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 kafka_server]# vim vars/main.yml
zk_cluster: 172.16.7.151:2181,172.16.7.152:2181,172.16.7.153:2181
kafka_manager_ip: 172.16.7.151 

另外在template的文件中还使用了个变量{{broker_id}}，该变量每台主机的值是不一样的，所以定义在了/etc/ansible/hosts文件中:

[kafka_servers]
172.16.206.17 broker_id=0
172.16.206.31 broker_id=1
172.16.206.32 broker_id=2

5. 设置主机组

/etc/ansible/hosts文件：

[kafka_servers]
172.16.206.17 broker_id=0
172.16.206.31 broker_id=1
172.16.206.32 broker_id=2

6. 使用角色

在roles同级目录，创建一个kafka.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim kafka.yml 
- hosts: kafka_servers
  remote_user: root
  roles:
  - kafka_server

运行playbook安装kafka集群：

[root@node1 ansible_playbooks]# ansible-playbook kafka.yml 

使用kafka_server role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.5 Elasticsearch role

[root@node1 es_server]# pwd
/root/ansible_playbooks/roles/es_server
[root@node1 es_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将elasticsearch-2.3.3.tar.gz  elasticsearch-analysis-ik-1.9.3.zip上传到files目录。

2. 编写tasks

[root@node1 es_server]# vim tasks/main.yml
- name: create es user
  user: name=es password={{password}}
  vars:
     # created with:
     # python -c 'import crypt; print crypt.crypt("This is my Password", "$1$SomeSalt$")'
     # >>> import crypt
     # >>> crypt.crypt('wisedu123', '$1$bigrandomsalt$')
     # '$1$bigrando$wzfZ2ifoHJPvaMuAelsBq0'
     password: $1$bigrando$wzfZ2ifoHJPvaMuAelsBq0
- name: mkdir directory for elasticsearch data
  file: dest=/esdata mode=0755 state=directory owner=es group=es
- name: copy and unzip es
  #unarchive module owner and group only effect on directory.
  unarchive: src=elasticsearch-2.3.3.tar.gz dest=/usr/local/
- name: install memory configuration file for es
  template: src=elasticsearch.in.sh.j2 dest=/usr/local/elasticsearch-2.3.3/bin/elasticsearch.in.sh owner=es group=es
- name: install configuration file for es
  template: src=elasticsearch.yml.j2 dest=/usr/local/elasticsearch-2.3.3/config/elasticsearch.yml owner=es group=es
- name: mkdir directory for elasticsearch-analysis-ik plugin
  file: dest=/usr/local/elasticsearch-2.3.3/plugins/ik mode=0755 state=directory owner=es group=es
- name: copy and unizp elasticsearch-analysis-ik plugin
  unarchive: src=elasticsearch-analysis-ik-1.9.3.zip dest=/usr/local/elasticsearch-2.3.3/plugins/ik
- name: change owner and group
  #recurse=yes make all files in a directory changed.
  file: path=/usr/local/elasticsearch-2.3.3 owner=es group=es recurse=yes
- name: start es
  shell: su - es -c '/usr/local/elasticsearch-2.3.3/bin/elasticsearch -d'
  #command: /usr/local/elasticsearch-2.3.3/bin/elasticsearch -d
  #become: true
  #become_method: su
  #become_user: es
  tags:
  - start

3. 编写templates

将模板elasticsearch.in.sh.j2和elasticsearch.yml.j2放入templates目录下

注意模板里的变量名中间不能用.。比如：{{node.name}}这样的变量名是不合法的。

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不再解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 es_server]# vim vars/main.yml
ES_MEM: 2g
cluster_name: wisedu
master_ip: 172.16.7.151 

另外在template的文件中还使用了个变量{{node_master}}，该变量每台主机的值是不一样的，所以定义在了/etc/ansible/hosts文件中:

[es_servers]
172.16.7.151 node_master=true
172.16.7.152 node_master=false
172.16.7.153 node_master=false 

5. 设置主机组

/etc/ansible/hosts文件：

[es_servers]
172.16.7.151 node_master=true
172.16.7.152 node_master=false
172.16.7.153 node_master=false 

6. 使用角色

在roles同级目录，创建一个es.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim es.yml 
- hosts: es_servers
  remote_user: root
  roles:
  - es_server

运行playbook安装Elasticsearch集群：

[root@node1 ansible_playbooks]# ansible-playbook es.yml

使用es_server role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.6 MySQL role

[root@node1 db_server]# pwd
/root/ansible_playbooks/roles/db_server
[root@node1 db_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将制作好的rpm包mysql-5.6.27-1.x86_64.rpm放到/root/ansible_playbooks/roles/db_server/files/目录下。
【注意】:这个rpm包是自己打包制作的，打包成rpm会使得部署的效率提高。关于如何打包成rpm见之前的博客《速成RPM包制作》。

2. 编写tasks

[root@node1 db_server]# vim tasks/main.yml
- name: install dependency package
  yum: name={{ item }} state=present
  with_items:
  - libaio
  - libaio-devel
- name: copy mysql rpm
  copy: src=mysql-5.6.27-1.x86_64.rpm dest=/tmp/
- name: install mysql
  yum: name=/tmp/mysql-5.6.27-1.x86_64.rpm state=present
- name: start mysql
  shell: /etc/init.d/mysqld start
  tags:
  - start
- name: set up root password
  shell: /usr/local/mysql/bin/mysql -uroot -e "UPDATE mysql.user SET Password=PASSWORD('wisedu123') where USER='root'" &>/dev/null
- name: delete anonymous account1
  shell: /usr/local/mysql/bin/mysql -uroot -Dmysql -pwisedu123 -e "DROP USER ''@localhost" &>/dev/null
- name: delete anonymous account2
  shell: /usr/local/mysql/bin/mysql -uroot -Dmysql -pwisedu123 -e "grant all on *.* to root@'%.%.%.%' identified by 'wisedu123'" &>/dev/null
- name: flush privileges
  shell: /usr/local/mysql/bin/mysql -uroot -Dmysql -pwisedu123 -e "flush privileges" &>/dev/null

3. 设置主机组

# vim /etc/ansible/hosts 
[db_servers]
172.16.7.152 

4. 使用角色

在roles同级目录，创建一个db.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim db.yml 
- hosts: mysql_server
  remote_user: root
  roles:
  - db_server

运行playbook安装MySQL：

[root@node1 ansible_playbooks]# ansible-playbook db.yml 

使用db_server role需要根据实际环境修改/etc/ansible/hosts文件里定义的主机。

3.7 Nginx role

[root@node1 web_server]# pwd
/root/ansible_playbooks/roles/web_server
[root@node1 web_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将制作好的rpm包openresty-for-godseye-1.9.7.3-1.x86_64.rpm放到/root/ansible_playbooks/roles/web_server/files/目录下。
【注意】:做成rpm包，在安装时省去了编译nginx的过程，提升了部署效率。这个包里面打包了很多与我们系统相关的文件。

2. 编写tasks

[root@node1 web_server]# vim tasks/main.yml 
- name: install dependency package
  yum: name={{ item }} state=present
  with_items:
  - openssl-devel
  - readline-devel
  - pcre-devel
  - gcc
- name: copy nginx
  copy: src=openresty-for-godseye-1.9.7.3-1.x86_64.rpm dest=/tmp/
- name: install nginx
  yum: name=/tmp/openresty-for-godseye-1.9.7.3-1.x86_64.rpm state=present
- name: install configuration file for nginx
  template: src=nginx.conf.j2 dest=/usr/local/openresty/nginx/conf/nginx.conf
- name: crontab task
  cron: name="clear nginx logs" weekday="0" hour="0" minute="0" job="/usr/local/openresty/clrnginxlog.sh"
- name: start nginx
  shell: systemctl start nginx.service
  tags:
  - start

3. 编写templates

将模板nginx.conf.j2放入templates目录下.

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不再解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 web_server]# vim vars/main.yml 
elasticsearch_cluster: server 172.16.7.151:9200;server 172.16.7.152:9200;server 172.16.7.153:9200;
kafka_server1: 172.16.7.151
kafka_server2: 172.16.7.152
kafka_server3: 172.16.7.153 

经过测试，变量里面不能有逗号。

5. 设置主机组

/etc/ansible/hosts文件：

# vim /etc/ansible/hosts 
[nginx_servers]
172.16.7.153

6. 使用角色

在roles同级目录，创建一个nginx.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim nginx.yml 
- hosts: nginx_servers
  remote_user: root
  roles:
  - web_server

运行playbook安装Nginx：

[root@node1 ansible_playbooks]# ansible-playbook nginx.yml

使用web_server role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.8 Redis role

[root@node1 redis_server]# pwd
/root/ansible_playbooks/roles/redis_server
[root@node1 redis_server]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将制作好的rpm包redis-3.2.2-1.x86_64.rpm放到/root/ansible_playbooks/roles/redis_server/files/目录下。

2. 编写tasks

[root@node1 redis_server]# vim tasks/main.yml
- name: install dependency package
  yum: name={{ item }} state=present
  with_items:
  - openssl-devel
  - readline-devel
  - pcre-devel
- name: copy redis
  copy: src=redis-3.2.2-1.x86_64.rpm dest=/tmp/
- name: install redis
  yum: name=/tmp/redis-3.2.2-1.x86_64.rpm state=present
- name: start redis
  shell: /usr/local/bin/redis-server /etc/redis.conf
  tags:
  - start

3. 设置主机组

/etc/ansible/hosts文件：

# vim /etc/ansible/hosts 
[redis_servers]
172.16.7.152

4. 使用角色

在roles同级目录，创建一个redis.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim redis.yml 
- hosts: redis_servers
  remote_user: root
  roles:
  - redis_server 

运行playbook安装redis：

[root@node1 ansible_playbooks]# ansible-playbook redis.yml 

使用redis_server role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.9 Hadoop role

完全分布式集群部署，NameNode和ResourceManager高可用。
提前配置集群节点的/etc/hosts文件、节点时间同步、某些集群主节点登录其他节点不需要输入密码。

[root@node1 hadoop]# pwd
/root/ansible_playbooks/roles/hadoop
[root@node1 hadoop]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将hadoop-2.7.2.tar.gz放到/root/ansible_playbooks/roles/hadoop/files/目录下。

2. 编写tasks

- name: install dependency package
  yum: name={{ item }} state=present
  with_items:
  - openssh
  - rsync
- name: create hadoop user
  user: name=hadoop password={{password}}
  vars:
    # created with:
    # python -c 'import crypt; print crypt.crypt("This is my Password", "$1$SomeSalt$")'
    # >>> import crypt
    # >>> crypt.crypt('wisedu123', '$1$bigrandomsalt$')
    # '$1$bigrando$wzfZ2ifoHJPvaMuAelsBq0'
    password: $1$bigrando$wzfZ2ifoHJPvaMuAelsBq0
- name: copy and unzip hadoop
  #unarchive module owner and group only effect on directory.
  unarchive: src=hadoop-2.7.2.tar.gz dest=/usr/local/
- name: create hadoop soft link
  file: src=/usr/local/hadoop-2.7.2 dest=/usr/local/hadoop state=link
- name: create hadoop logs directory
  file: dest=/usr/local/hadoop/logs mode=0775 state=directory
- name: change hadoop soft link owner and group
  #recurse=yes make all files in a directory changed.
  file: path=/usr/local/hadoop owner=hadoop group=hadoop recurse=yes
- name: change hadoop-2.7.2 directory owner and group
  #recurse=yes make all files in a directory changed.
  file: path=/usr/local/hadoop-2.7.2 owner=hadoop group=hadoop recurse=yes
- name: set hadoop env
  lineinfile: dest={{env_file}} insertafter="{{item.position}}" line="{{item.value}}" state=present
  with_items:
  - {position: EOF, value: "
"}
  - {position: EOF, value: "# Hadoop environment"}
  - {position: EOF, value: "export HADOOP_HOME=/usr/local/hadoop"}
  - {position: EOF, value: "export PATH=$PATH:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin"}
- name: enforce env
  shell: source {{env_file}}
- name: install configuration file hadoop-env.sh.j2 for hadoop
  template: src=hadoop-env.sh.j2 dest=/usr/local/hadoop/etc/hadoop/hadoop-env.sh owner=hadoop group=hadoop
- name: install configuration file core-site.xml.j2 for hadoop
  template: src=core-site.xml.j2 dest=/usr/local/hadoop/etc/hadoop/core-site.xml owner=hadoop group=hadoop
- name: install configuration file hdfs-site.xml.j2 for hadoop
  template: src=hdfs-site.xml.j2 dest=/usr/local/hadoop/etc/hadoop/hdfs-site.xml owner=hadoop group=hadoop
- name: install configuration file mapred-site.xml.j2 for hadoop
  template: src=mapred-site.xml.j2 dest=/usr/local/hadoop/etc/hadoop/mapred-site.xml owner=hadoop group=hadoop
- name: install configuration file yarn-site.xml.j2 for hadoop
  template: src=yarn-site.xml.j2 dest=/usr/local/hadoop/etc/hadoop/yarn-site.xml owner=hadoop group=hadoop
- name: install configuration file slaves.j2 for hadoop
  template: src=slaves.j2 dest=/usr/local/hadoop/etc/hadoop/slaves owner=hadoop group=hadoop
- name: install configuration file hadoop-daemon.sh.j2 for hadoop
  template: src=hadoop-daemon.sh.j2 dest=/usr/local/hadoop/sbin/hadoop-daemon.sh owner=hadoop group=hadoop
- name: install configuration file yarn-daemon.sh.j2 for hadoop
  template: src=yarn-daemon.sh.j2 dest=/usr/local/hadoop/sbin/yarn-daemon.sh owner=hadoop group=hadoop
# make sure zookeeper started, and then start hadoop.
# start journalnode
- name: start journalnode
  shell: /usr/local/hadoop/sbin/hadoop-daemon.sh start journalnode
  become: true
  become_method: su
  become_user: hadoop
  when: datanode == "true"
# format namenode
- name: format active namenode hdfs
  shell: /usr/local/hadoop/bin/hdfs namenode -format
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
- name: start active namenode hdfs
  shell: /usr/local/hadoop/sbin/hadoop-daemon.sh start namenode
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
- name: format standby namenode hdfs
  shell: /usr/local/hadoop/bin/hdfs namenode -bootstrapStandby
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_standby == "true"
- name: stop active namenode hdfs
  shell: /usr/local/hadoop/sbin/hadoop-daemon.sh stop namenode
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
# format ZKFC
- name: format ZKFC
  shell: /usr/local/hadoop/bin/hdfs zkfc -formatZK
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
# start hadoop cluster
- name: start namenode
  shell: /usr/local/hadoop/sbin/start-dfs.sh
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
- name: start yarn
  shell: /usr/local/hadoop/sbin/start-yarn.sh
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_active == "true"
- name: start standby rm
  shell: /usr/local/hadoop/sbin/yarn-daemon.sh start resourcemanager
  become: true
  become_method: su
  become_user: hadoop
  when: namenode_standby == "true" 

3. 编写templates

将模板core-site.xml.j2、hadoop-daemon.sh.j2、hadoop-env.sh.j2、hdfs-site.xml.j2、mapred-site.xml.j2、slaves.j2、yarn-daemon.sh.j2、yarn-site.xml.j2放入templates目录下。

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不再解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 hadoop]# vim vars/main.yml 
env_file: /etc/profile
# hadoop-env.sh.j2 file variables.
JAVA_HOME: /usr/java/jdk1.8.0_73
# core-site.xml.j2 file variables.
ZK_NODE1: node1:2181
ZK_NODE2: node2:2181
ZK_NODE3: node3:2181
# hdfs-site.xml.j2 file variables.
NAMENODE1_HOSTNAME: node1
NAMENODE2_HOSTNAME: node2
DATANODE1_HOSTNAME: node3
DATANODE2_HOSTNAME: node4
DATANODE3_HOSTNAME: node5
# mapred-site.xml.j2 file variables.
MR_MODE: yarn
# yarn-site.xml.j2 file variables.
RM1_HOSTNAME: node1
RM2_HOSTNAME: node2

5. 设置主机组

/etc/ansible/hosts文件：

# vim /etc/ansible/hosts 
[hadoop]
172.16.7.151 namenode_active=true namenode_standby=false datanode=false
172.16.7.152 namenode_active=false namenode_standby=true datanode=false
172.16.7.153 namenode_active=false namenode_standby=false datanode=true
172.16.7.154 namenode_active=false namenode_standby=false datanode=true
172.16.7.155 namenode_active=false namenode_standby=false datanode=true

6. 使用角色

在roles同级目录，创建一个hadoop.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim hadoop.yml 
- hosts: hadoop
  remote_user: root
  roles:
  - jdk8
  - hadoop

运行playbook安装hadoop集群：

[root@node1 ansible_playbooks]# ansible-playbook hadoop.yml 

使用hadoop role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

3.10 Spark role

Standalone模式部署spark (无HA)

[root@node1 spark]# pwd
/root/ansible_playbooks/roles/spark
[root@node1 spark]# ls
files  handlers  meta  tasks  templates  vars

1. 上传安装包

将scala-2.10.6.tgz和spark-1.6.1-bin-hadoop2.6.tgz放到/root/ansible_playbooks/roles/hadoop/files/目录下。

2. 编写tasks

- name: copy and unzip scala
  unarchive: src=scala-2.10.6.tgz dest=/usr/local/
- name: set scala env
  lineinfile: dest={{env_file}} insertafter="{{item.position}}" line="{{item.value}}" state=present
  with_items:
  - {position: EOF, value: "
"}
  - {position: EOF, value: "# Scala environment"}
  - {position: EOF, value: "export SCALA_HOME=/usr/local/scala-2.10.6"}
  - {position: EOF, value: "export PATH=$SCALA_HOME/bin:$PATH"}
- name: copy and unzip spark
  unarchive: src=spark-1.6.1-bin-hadoop2.6.tgz dest=/usr/local/
- name: rename spark directory
  command: mv /usr/local/spark-1.6.1-bin-hadoop2.6 /usr/local/spark-1.6.1
- name: set spark env
  lineinfile: dest={{env_file}} insertafter="{{item.position}}" line="{{item.value}}" state=present
  with_items:
  - {position: EOF, value: "
"}
  - {position: EOF, value: "# Spark environment"}
  - {position: EOF, value: "export SPARK_HOME=/usr/local/spark-1.6.1"}
  - {position: EOF, value: "export PATH=$SPARK_HOME/bin:$PATH"}
- name: enforce env
  shell: source {{env_file}}
- name: install configuration file for spark
  template: src=slaves.j2 dest=/usr/local/spark-1.6.1/conf/slaves
- name: install configuration file for spark
  template: src=spark-env.sh.j2 dest=/usr/local/spark-1.6.1/conf/spark-env.sh
- name: start spark cluster
  shell: /usr/local/spark-1.6.1/sbin/start-all.sh
  tags:
  - start

3. 编写templates

将模板slaves.j2和spark-env.sh.j2放到/root/ansible_playbooks/roles/spark/templates/目录下。

配置文件内容过多，具体见github，地址是https://github.com/jkzhao/ansible-godseye。配置文件内容也不再解释，在前面博客中的文章中都已写明。

4. 编写vars

[root@node1 spark]# vim vars/main.yml 
env_file: /etc/profile
# spark-env.sh.j2 file variables
JAVA_HOME: /usr/java/jdk1.8.0_73
SCALA_HOME: /usr/local/scala-2.10.6
SPARK_MASTER_HOSTNAME: node1
SPARK_HOME: /usr/local/spark-1.6.1
SPARK_WORKER_MEMORY: 256M
HIVE_HOME: /usr/local/apache-hive-2.1.0-bin
HADOOP_CONF_DIR: /usr/local/hadoop/etc/hadoop/
# slave.j2 file variables
SLAVE1_HOSTNAME: node2
SLAVE2_HOSTNAME: node3

5. 设置主机组

/etc/ansible/hosts文件：

# vim /etc/ansible/hosts 
[spark]
172.16.7.151
172.16.7.152
172.16.7.153

6. 使用角色

在roles同级目录，创建一个spark.yml文件，里面定义好你的playbook。

[root@node1 ansible_playbooks]# vim spark.yml 
- hosts: spark
  remote_user: root
  roles:
  - spark 

运行playbook安装spark集群：

[root@node1 ansible_playbooks]# ansible-playbook spark.yml 

使用spark role需要根据实际环境修改vars/main.yml里的变量以及/etc/ansible/hosts文件里定义的主机。

【注】：所有的文件都在github上，https://github.com/jkzhao/ansible-godseye。