ansible hosts及Inventory文件

主机目录配置文件

  • 默认文件 : /etc/ansible/hosts

修改主机目录的配置文件: /etc/ansible/ansible.cfg, 指定inventory的路径

1
inventory      = /data/db/playbook/inventory
  • 命令行中传递主机目录配置文件。ansible-playbook -i inventory site.ymlansible-playbook --inventory-file inventory site.yml

如何编写 Hosts and Groups(主机与组)

简单的主机和组

1
2
3
4
5
6
7
8
9
10
mail.example.com

[webservers]
foo.example.com
bar.example.com

[dbservers]
one.example.com
two.example.com
three.example.com
  • 中括号中的名字代表组名,可以根据自己的需求将庞大的主机分成具有标识的组,如上面分了两个组webservers和dbservers组。
  • 主机(hosts)部分可以使用域名、主机名、IP地址表示;当然使用前两者时,也需要主机能反解析到相应的IP地址,一般此类配置中多使用IP地址。

单独配置端口与别名

如果某些主机的SSH运行在自定义的端口上,ansible使用Paramiko进行ssh连接时,不会使用你SSH配置文件中列出的端口,但是如果修改ansible使用openssh进行ssh连接时将会使用:

1
2
[webservers]
192.168.56.13:52113

假如你想要为某些静态IP设置一些别名,可以这样做:

1
test1 ansible_ssh_port = 52113 ansible_ssh_host = 192.168.56.13

上面的 test1别名就指代了IP为192.168.56.13,ssh连接端口为52113的主机。

使用正则指定主机范围

1
2
3
4
5
[webservers]
www[01:50].example.com

[databases]
db-[a:f].example.com

上面指定了从web1到web50,webservers组共计50台主机;databases组有db-a到db-f共6台主机。

使用主机变量

以下是Hosts部分中经常用到的变量部分:

  • ansible_ssh_host: #用于指定被管理的主机的真实IP。
  • ansible_ssh_port: #用于指定连接到被管理主机的ssh端口号,默认是22。
  • ansible_ssh_user: #ssh连接时默认使用的用户名。
  • ansible_ssh_pass: #ssh连接时的密码。
  • ansible_sudo_pass: #使用sudo连接用户时的密码。
  • ansible_sudo_exec: #如果sudo命令不在默认路径,需要指定sudo命令路径。
  • ansible_ssh_private_key_file: #秘钥文件路径,秘钥文件如果不想使用ssh-agent管理时可以使用此选项。
  • ansible_shell_type: #目标系统的shell的类型,默认sh。
  • ansible_connection: #SSH 连接的类型: local , ssh , paramiko,在 ansible 1.2 之前默认是 paramiko ,后来智能选择,优先使用基于 ControlPersist 的 ssh (支持的前提)。
  • ansible_python_interpreter: #用来指定python解释器的路径,默认为/usr/bin/python 同样可以指定ruby 、perl 的路径。
  • ansible_*_interpreter: #其他解释器路径,用法与ansible_python_interpreter类似,这里”*”可以是ruby或才perl等其他语言。

示例如下:

1
2
3
4
[test]
192.168.1.1 ansible_ssh_user=root ansible_ssh_pass='abcdefg'
192.168.1.2 ansible_ssh_user=breeze ansible_ssh_pass='123456'
192.168.1.3 ansible_ssh_user=bernie ansible_ssh_port=3055 ansible_ssh_pass='456789'

上面的示例中指定了三台主机,三台主机的用密码分别是abcdefg、123456、45789,指定的ssh连接的用户名分别为root、breeze、bernie,ssh 端口分别为22、22、3055 ,这样在ansible命令执行的时候就不用再指令用户和密码等了。

组内变量

变量也可以通过组名,应用到组内的所有成员:

1
2
3
4
5
6
7
[webserver]
host1
host2

[webserver:vars]
ntp_server=ntp.atlanta.example.com
proxy=proxy.atlanta.example.com

上面test组中包含两台主机,通过对test组指定vars变更,相应的host1和host2相当于相应的指定了ntp_server和proxy变量参数值 。

按目录结构存储变量

假设inventory文件为/etc/ansible/hosts,那么相关的hosts和group变量可以放在下面的目录结构下

1
2
3
/etc/ansible/group_vars/webserver # can optionally end in '.yml', '.yaml', or '.json'
/etc/ansible/group_vars/webservers
/etc/ansible/host_vars/foosball

/etc/ansible/group_vars/webserver文件内容可以为

1
2
3
---
ntp_server: acme.example.org
database_server: storage.example.org

如果对应的名字为目录名,ansible会读取这个目录下面所有文件的内容

1
2
/etc/ansible/group_vars/raleigh/db_settings
/etc/ansible/group_vars/raleigh/cluster_settings

group_vars/ 和 host_vars/ 目录可放在 inventory 目录下,或是 playbook 目录下. 如果两个目录下都存在,那么 playbook 目录下的配置会覆盖 inventory 目录的配置.

组的包含与组内变量

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
[atlanta]
host1
host2

[raleigh]
host2
host3

[southeast:children]
atlanta
raleigh

[southeast:vars]
some_server=foo.southeast.example.com
halon_system_timeout=30
self_destruct_countdown=60
escape_pods=2

[usa:children]
southeast
northeast
southwest
northwest
  • 任何属于子组的成员都自动成为父组的成员。
  • 子组的变量将具有更高的优先级(覆盖)父组的变量。
  • 组可以有多个父母和孩子,但不是循环关系。
  • 主机也可以在多个组中,但只有一个主机实例,合并来自多个组的数据。

注:vars变量在ansible ad-hoc部分中基本用不到,主要用在ansible-playbook中。

Patterns(主机与组正则匹配部分)

把Patterns 直接理解为正则实际是不完全准确的,正常的理解为patterns意味着在ansible中管理哪些主机,也可以理解为,要与哪台主机进行通信。在探讨这个问题之前我们先看下ansible的用法:

1
ansible <pattern_goes_here> -m <module_name> -a <arguments>

直接上一个示例:

1
ansible webservers -m service -a "name=httpd state=restarted"

这里是对webservers 组或主机重启httpd服务 ,其中webservers 就是Pattern部分。而之所以上面说Pattern(模式)可以理解为正则,主要针对下面经常用到的用法而言的。

  1. 表示所有的主机可以使用all 或 *
  2. 通配符与逻辑或
  3. 利用通配符还可以指定一组具有规则特征的主机或主机名,冒号表示or---逻辑或
1
2
3
4
web1.biglittleant.com
web1.biglittleant.com:web2.biglittleant.com
192.168.1.1
192.168.1.*

当然,这里的*通配符也可以用在前面,如:

1
2
3
4
*.biglittleant.com
*.com
webservers1[0] #表示匹配 webservers1 组的第 1 个主机
webservers1[0:25] #表示匹配 webservers1 组的第 1 个到第 25 个主机(官网文档是":"表示范围,测试发现应该使用"-",注意不要和匹配多个主机组混淆)

上面的用法,在多个组之间同样适用 ,如:

1
2
webservers
webservers:dbservers #表示两个组中所有的主机

逻辑非与逻辑and

非的表达式,如,目标主机必须在组webservers但不在phoenix组中

1
webserver:!phoenix

交集的表达式,如,目标主机必须即在组webservers中又在组staging中

1
webservers:&staging

一个更复杂的示例:

1
webserver:dbservers:&staging:!phoenix

上面这个复杂的表达式最后表示的目标主机必须满足:在webservers或者dbservers组中,必须还存在于staging组中,但是不在phoenix组中 。

混合高级用法

1
*.yanruogu.com:*.org

还可以在开头的地方使用”~”,用来表示这是一个正则表达式:

1
~(web|db).*\.yanruogu\.com

给两个ansible-playbook中具体可能用的用法:

  1. 在ansible-palybook命令中,你也可以使用变量来组成这样的表达式,但是你必须使用“-e”的选项来指定这个表达式(通常我们不这样用):
1
ansible-palybook -e webservers:!`excluded`:&`required`
  1. 在ansible和ansible-playbook中,还可以通过一个参数”–limit”来明确指定排除某些主机或组:
1
ansible-playbook site.yml --limit datacenter2
  1. 从Ansible1.2开始,如果想排除一个文件中的主机可以使用”@”:
1
ansible-playbook site.yml --limit @retry_hosts.txt

动态 定义主机组

踩坑

1
ansible -i inventory/ other -m ping

当执行如上命令时,other 如果有多个组,他会将所有的other组都执行。而不是报错。

参考文档

Working with Inventory