在大量节点加入Puppet之后，你至少会面临两个比较大的问题：
1、由于节点数的增多，site.pp文件必然会编写更多的节点条目，以及节点包含的类。假设你用Puppet管理500个节点，存在三种情况：1、所有节点有共同的类，也可以理解为模块；2、所有节点分成了50组，每组节点有不同的应用，每组应用编写一个模块；3、每个节点也应该有一个自己的模块，在这种环境下，可想而知，site.pp文件是多么的复杂，多么的难以维护。
2、在使用MCO推送更新的时候，后面可以跟上fact变量用于过滤一组节点。而官方给的facter只能满足普遍存在的fact变量，可是，你要对一组节点进行归类怎么办呢，比如有10个节点的应用是Apache，15个节点属于Nginx等，那么就要自定义fact变量。现在面临的问题是，所有节点都要自定义fact变量，又如何做到集中管理呢？
其实、节点与类之间的勾连，更多时候是要借助ENC（外部节点分类器）来完成，因为puppet软件本身并不具备此类功能。Puppet-dashboard企业版和Foreman已经具备了这方面的功能，而且是图形界面操作，还是很方便的。众所周知，Puppet-dashboard社区版目前最新版本为1.2.23，到目前为止已经有一年多未更新了，可见Puppetlabs的意图很明确，逼着你用企业版。而Foreman部署和使用略复杂些，其次Foreman很重，它的内部并不仅仅是包含puppet还包含了很多其它管理工具，非常庞大，可以酌情考虑使用。接下来要介绍的是官方推荐的hiera软件，一个很强大的ENC，编写的文件只要遵循yaml格式即可，使用非常方便，可惜的是没有图形界面。但这并不代表你不会开发图形界面哦。
hiera工具和Puppetmaster结合需要通过hiera-puppet工具完成，在puppet-server 3.0.*版本后，这个工具已经集成到了puppet-server安装包中。而我们现在使用的环境仍然是puppet2.7版本，所以需要安装hiera-puppet工具。
[root@puppetmaster RHEL6U4]# yum install hiera hiera-puppet #注意依赖关系
...
Dependencies Resolved
===========================================================================
Package Arch Version Repository Size
===========================================================================
Installing:
hiera noarch 1.3.2-1.el6 rhel-puppet 23 k
hiera-puppet noarch 1.0.0-1.el6 rhel-puppet 14 k
Installing for dependencies:
ruby-irb x86_64 1.8.7.352-7.el6_2 rhel-base 311 k
ruby-rdoc x86_64 1.8.7.352-7.el6_2 rhel-base 375 k
rubygem-json x86_64 1.5.5-1.el6 rhel-puppet 763 k
rubygems noarch 1.3.7-1.el6 rhel-puppet 206 k
Transaction Summary
===========================================================================
...默认hiera.yaml主配置文件在/etc目录下，为了结合后期版本控制系统集中管理，建议将此文件copy到/etc/puppet目录下，然后创建软连接指向/etc/hiera.yaml即可。
[root@puppetmaster ~]# mv /etc/hiera.yaml /etc/puppet/
[root@puppetmaster ~]# ln -s /etc/puppet/hiera.yaml /etc/hiera.yaml
[root@puppetmaster ~]# ll /etc/hiera.yaml
lrwxrwxrwx 1 root root 22 Apr 20 20:05 /etc/hiera.yaml -> /etc/puppet/hiera.yaml备注： 以上变量其实就是fact变量。
[root@puppetmaster ~]# vim /etc/puppet/hiera.yaml
---
:backends:
- yaml
:hierarchy:
# - defaults
# - "%{clientcert}"
- common
- "%{environment}"
- "%{osfamily}"
- "%{hostname}"
# - global
:yaml:
:datadir:"/etc/puppet/environments/%{environment}/hiera"hiera主配置文件编写完成之后，需要重启puppetmaster后方可生效。
[root@puppetmaster hiera]# /etc/init.d/puppetmaster restart
Stopping puppetmaster: [ OK ]
Starting puppetmaster: [ OK ][root@puppetmaster hiera]# vim common.yaml
---
puppetserver:
- 'puppetmaster.kisspuppet.com'通过hiera命令测试
[root@puppetmaster ~]# hiera puppetserver environment=kissprd
["puppetmaster.kisspuppet.com"]
[root@puppetmaster ~]# hiera puppetserver environment=kissdev
nil通过以上命令可以得知在环境为kissprd的情况下，puppetserver的变量为puppetmaster.kisspuppet.com，值为nil的意思是不存在。
[root@agent1 ~]# facter osfamily
RedHat
[root@puppetmaster hiera]# vim RedHat.yaml
---
classes:
- 'puppet'
- 'yum'通过hiera命令测试
[root@puppetmaster hiera]# hiera classes environment=kissprd
nil
[root@puppetmaster hiera]# hiera classes environment=kissprd osfamily=RedHat
["motd", "puppet", "yum"]
[root@puppetmaster hiera]# hiera classes environment=kissprd osfamily=SLES
nil通过以上命令可以得在环境为kissprd，系统为RedHat的情况下，classes的变量为三个值（puppet、yum）。
[root@agent1 ~]# facter hostname
agent1
[root@puppetmaster hiera]# vim agent1.yaml
---
classes:
- 'motd'
certname:
- 'agent1_cert.kisspuppet.com'
[root@puppetmaster hiera]# vim agent2.yaml
---
classes:
- 'motd'
certname:
- 'agent2_cert.kisspuppet.com'
[root@puppetmaster hiera]# vim agent3.yaml
---
certname:
- 'agent3_cert.kisspuppet.com'通过hiera命令测试
[root@puppetmaster hiera]# hiera classes environment=kissprd hostname=agent
1
["motd"]
[root@puppetmaster hiera]# hiera classes environment=kissprd hostname=agent
2
["motd"]
[root@puppetmaster hiera]# hiera classes environment=kissprd hostname=agent
3
nil
[root@puppetmaster hiera]# hiera certname environment=kissprd hostname=agent1
["agent1_cert.kisspuppet.com"]
[root@puppetmaster hiera]# hiera certname environment=kissprd hostname=agent2
["agent2_cert.kisspuppet.com"]
[root@puppetmaster hiera]# hiera certname environment=kissprd hostname=agent3
["agent3_cert.kisspuppet.com"]通过以上命令测试可以得知，系统fact变量hostname为agent1和agent2的情况下，hiera变量classes为motd。certname变量为各自的certname变量。
以前的写法：
[root@puppetmaster ~]# vim /etc/puppet/environments/kissprd/manifests/site.pp
$puppetserver = 'puppetmaster.kisspuppet.com'
class environments{
include puppet,yum
}
node 'puppetmaster_cert.kisspuppet.com'{
include environments
}
node 'agent1_cert.kisspuppet.com'{
include environments,motd
}
node 'agent2_cert.kisspuppet.com'{
include environments,motd
}
node 'agent3_cert.kisspuppet.com'{
include environments
}应用了hiera之后的写法：
[root@puppetmaster ~]# vim /etc/puppet/environments/kissprd/manifests/site.pp
$puppetserver = hiera('puppetserver') #引用了hiera中common.yaml中的全局变量puppetserver
node default {
hiera_include('classes') #引用了hiera中osfamily和hostname变量
}备注：以后添加节点，只需要编写yaml文件即可，而site.pp文件无需在进行修改。
未使用hiera之前的编写方式
[root@puppetmaster hiera]# vim /etc/puppet/environments/kissprd/environment/modules/puppet/manifests/params.pp
class puppet::params {
$puppetserver = 'puppetmaster.kisspuppet.com'
$certname = "${::hostname}_cert.kisspuppet.com"
case $operatingsystemmajrelease{
5: {
$puppet_release = '2.7.23-1.el5'
$facter_release = '1.7.3-1.el5'
}
6: {
$puppet_release = '2.7.23-1.el6'
$facter_release = '1.7.3-1.el6'
}
default: {
fail("Module puppet is not supported on ${::operatingsystem}")
}
}
}应用了hiera变量之后的编写方式
[root@puppetmaster hiera]# vim /etc/puppet/environments/kissprd/environment/modules/puppet/manifests/params.pp
class puppet::params {
$puppetserver = hiera('puppetserver') #应用了hiera全局文件common.yaml中的puppetserver变量
$certname = hiera('certname') #应用了hiera中变量hostname对应的节点文件$hostname.yaml中的certname变量
case $operatingsystemmajrelease{
5: {
$puppet_release = '2.7.23-1.el5'
$facter_release = '1.7.3-1.el5'
}
6: {
$puppet_release = '2.7.23-1.el6'
$facter_release = '1.7.3-1.el6'
}
default: {
fail("Module puppet is not supported on ${::operatingsystem}")
}
}
}[root@agent1 ~]# puppet agent -t --environment=kissprd
info: Retrieving plugin
info: Loading facts in /var/lib/puppet/lib/facter/hwclock.rb
info: Caching catalog for agent1_cert.kisspuppet.com
info: Applying configuration version '1398007834'
notice: Hardware-Clock: Sun 20 Apr 2014 11:30:35 PM CST -0.059676 seconds
notice: /Stage[main]/Motd/Notify[ Hardware-Clock: Sun 20 Apr 2014 11:30:35 PM CST -0.059676 seconds]/message: defined 'message' as ' Hardware-Clock: Sun 20 Apr 2014 11:30:35 PM CST -0.059676 seconds'
notice: /Stage[main]/Motd/File[/etc/motd]/content:
--- /etc/motd 2014-04-20 23:30:26.000000000 +0800
+++ /tmp/puppet-file20140420-9781-fm0gig-0 2014-04-20 23:30:37.000000000 +0800
@@ -0,0 +1,3 @@
+-- --
+--------puppet test---------
+-- --
info: FileBucket got a duplicate file {md5}d41d8cd98f00b204e9800998ecf8427e
info: /Stage[main]/Motd/File[/etc/motd]: Filebucketed /etc/motd to puppet with sum d41d8cd98f00b204e9800998ecf8427e
notice: /Stage[main]/Motd/File[/etc/motd]/content: content changed '{md5}d41d8cd98f00b204e9800998ecf8427e' to '{md5}87ea3a1af8650395038472457cc7f2b1'
notice: Finished catalog run in 0.76 seconds
[root@agent1 ~]# cat /etc/motd
-- --
--------puppet test---------
-- --
[root@agent3 ~]# >/etc/motd
[root@agent3 ~]# puppet agent -t --environment=kissprd
info: Retrieving plugin
notice: /File[/var/lib/puppet/lib/facter]/ensure: created
notice: /File[/var/lib/puppet/lib/facter/hwclock.rb]/ensure: defined content as '{md5}d8cc9fe2b349a06f087692763c878e28'
info: Loading downloaded plugin /var/lib/puppet/lib/facter/hwclock.rb
info: Loading facts in /var/lib/puppet/lib/facter/hwclock.rb
info: Caching catalog for agent3_cert.kisspuppet.com
info: Applying configuration version '1398007834'
notice: Finished catalog run in 0.67 seconds
[root@agent3 ~]# cat /etc/motd
[root@agent3 ~]# 从以上测试可以得知类motd应用到了agent1并未应用到agent3，说明hiera传输变量正确。
[root@agent3 ~]# vim /etc/puppet/puppet.conf
### config by puppet ###
[main]
logdir = /var/log/puppet
rundir = /var/run/puppet
ssldir = $vardir/ssl
pluginsync = true
[agent]
classfile = $vardir/classes.txt
localconfig = $vardir/localconfig
# server = puppetmaster.kisspuppet.com #注释掉进行测试
# certname = agent3_cert.kisspuppet.com
runinterval = 10
[root@agent3 ~]# puppet agent -t --environment=kissprd --server=puppetmaster.kisspuppet.com --certname=agent3_cert.kisspuppet.com
info: Retrieving plugin
info: Loading facts in /var/lib/puppet/lib/facter/hwclock.rb
info: Caching catalog for agent3_cert.kisspuppet.com
info: Applying configuration version '1398008880'
notice: /Stage[main]/Puppet::Config/File[/etc/puppet/puppet.conf]/content:
--- /etc/puppet/puppet.conf 2014-04-20 23:46:05.839948111 +0800
+++ /tmp/puppet-file20140420-5618-45rytg-0 2014-04-20 23:48:01.653944253 +0800
@@ -7,6 +7,6 @@
[agent]
classfile = $vardir/classes.txt
localconfig = $vardir/localconfig
-# server = puppetmaster.kisspuppet.com
-# certname = agent3_cert.kisspuppet.com
+ server = puppetmaster.kisspuppet.com
+ certname = agent3_cert.kisspuppet.com
runinterval = 10
info: FileBucket adding {md5}13430e5962e7584c9422e5adc1f3ba43
info: /Stage[main]/Puppet::Config/File[/etc/puppet/puppet.conf]: Filebucketed /etc/puppet/puppet.conf to puppet with sum 13430e5962e7584c9422e5adc1f3ba43
notice: /Stage[main]/Puppet::Config/File[/etc/puppet/puppet.conf]/content: content changed '{md5}13430e5962e7584c9422e5adc1f3ba43' to '{md5}23545b7afd09af671920f122a20db952'
info: /Stage[main]/Puppet::Config/File[/etc/puppet/puppet.conf]: Scheduling refresh of Class[Puppet::Service]
info: Class[Puppet::Service]: Scheduling refresh of Service[puppet]
notice: /Service[puppet]: Triggered 'refresh' from 1 events
notice: Finished catalog run in 0.54 seconds
[root@agent3 ~]# cat /etc/puppet/puppet.conf
### config by puppet ###
[main]
logdir = /var/log/puppet
rundir = /var/run/puppet
ssldir = $vardir/ssl
pluginsync = true
[agent]
classfile = $vardir/classes.txt
localconfig = $vardir/localconfig
server = puppetmaster.kisspuppet.com
certname = agent3_cert.kisspuppet.com
runinterval = 10通过以上测试可以得知hiera变量certname和puppetserver传输正常。
特别说明： 由于hiera定义的变量需要通过转换才能变为puppet能够使用的变量，而这个转换是需要耗费CPU资源的，笔者曾经测试过一组节点（50个）同时传输hiera的变量数超过2000个，出现CPU负载过高的性能瓶颈问题需要特别注意。其次hiera数据除了用yaml格式保存外还可以存放在redis数据库中，这样查询起来性能会高很多，具体可参考官网，《pro puppet 》第二版也有这方面的介绍，可参阅。
通过以上测试至少解决了两个问题：
思路： 参考Puppet基础篇10-自定义fact实现的四种方式介绍在puppetmaster端编写一个facts模块，传输一个变量文件至节点对应的目录里面，变量文件名可以通过主机名进行定义。
之前facts模块中的结构
[root@puppetmaster modules]# tree facts/
facts/
├── files
├── lib
│ └── facter
│ └── hwclock.rb #通过pluginsync模式发布的自定义fact变量，无需修改
├── manifests
└── templates
5 directories, 1 file[root@puppetmaster manifests]# vim config.pp #定义file资源
class facts::config{
file{ "/etc/facter/facts.d/$hostname.txt": #文件名称通过变量hostname获取
owner => "root",
group => "root",
mode => 0400,
source => "puppet:///modules/facts/facts.d/$hostname.txt", #文件名称通过节点变量hostname获取
require => Class['facts::exec'],
}
}
[root@puppetmaster manifests]# vim exec.pp #定义可执行资源保证目录 /etc/facter/facts.d 存在
class facts::exec{
exec {"create fact external":
command => "mkdir -p /etc/facter/facts.d ",
path => ["/usr/bin","/usr/sbin","/bin","/sbin"],
creates => "/etc/facter/facts.d",
}
}
[root@puppetmaster manifests]# vim init.pp
class facts{
include facts::config,facts::exec
}[root@puppetmaster facts.d]# pwd
/etc/puppet/environments/kissprd/environment/modules/facts/files/facts.d
[root@puppetmaster facts.d]# vim agent1.txt
env=prd
app=weblogic
[root@puppetmaster facts.d]# vim agent2.txt
env=qa
app=db2
[root@puppetmaster facts.d]# vim agent3.txt
env=prd
app=nginx由于模块facts属于全局的，应用于common.ymal或者RedHat.ymal中即可。
[root@puppetmaster hiera]# vim RedHat.yaml
---
classes:
- 'puppet'
- 'yum'
- 'facts'[root@agent3 ~]# ll /etc/facter/facts.d
ls: cannot access /etc/facter/facts.d: No such file or directory
[root@agent3 ~]# puppet agent -t --environment=kissprd
info: Retrieving plugin
info: Loading facts in /var/lib/puppet/lib/facter/hwclock.rb
info: Caching catalog for agent3_cert.kisspuppet.com
info: Applying configuration version '1398010573'
notice: /Stage[main]/Facts::Exec/Exec[create fact external]/returns: executed successfully
notice: /Stage[main]/Facts::Config/File[/etc/facter/facts.d/agent3.txt]/ensure: defined content as '{md5}3330b8efe95f6747de47a9eca3a5411e'
notice: Finished catalog run in 0.66 seconds
[root@agent3 ~]# cat /etc/facter/facts.d/agent3.txt
env=prd
app=nginx
[root@agent3 ~]# facter env
prd
[root@agent3 ~]# facter app
nginx其它节点测试略
注意：以上方法只是提供给你一种集中管理自定义fact的思路，并不是最好的解决方案，只不过这种方法目前笔者用于上产环境中感觉还不错，特此分享。