MCollective可以使用多种方式进行扩展。最普遍的一种扩展MCollective的方式就是重用已经写好的agent插件。这些小的Ruby库可以让MCollective在整个集群中执行自定义的命令。
一个agent插件通常包含一个Ruby库，它必须被分发到所有运行MCollective agent的节点上。另外，一个数据定义文件（DDL）提供了插件接受的传入参数的具体描述，整个DDL文件需要放在MCollective客户端系统上。最后，一个使用指定的agent插件运行MCollective的脚步也需要被安装到所有的MCollective客户端系统上。
备注：更多插件可以在https://github.com/puppetlabs/mcollective-plugins找到。
1.1 安装puppet agent插件
MCollective本身并不包含一个可以立即使用的Puppet agent插件，需要安装使用。这一插件可以让操作员在需要时运行Puppet agent。他不需要等待Puppet agent的默认运行间隔，也不需要使用其他工具来开始这些任务
1.1.1 安装MCollective的Agent插件
[root@agent1 rpms]# yum install mcollective-puppet-agent mcollective-puppet-common
[root@agent1 rpms]# ll /usr/libexec/mcollective/mcollective/agent/
total 36
-rw-r--r-- 1 root root 1033 May 21 01:34 discovery.rb
-rw-r--r-- 1 root root 8346 May 14 07:28 puppet.ddl
-rw-r--r-- 1 root root 7975 May 14 07:25 puppet.rb
-rw-r--r-- 1 root root 5999 May 21 01:34 rpcutil.ddl
-rw-r--r-- 1 root root 3120 May 21 01:34 rpcutil.rb
[root@puppetserver rpms]# yum install mcollective-puppet-client mcollective-puppet-common
[root@puppetserver rpms]# ll /usr/libexec/mcollective/mcollective/agent/
total 28
-rw-r--r-- 1 root root 1033 May 21 01:34 discovery.rb
-rw-r--r-- 1 root root 8346 May 14 07:28 puppet.ddl
-rw-r--r-- 1 root root 5999 May 21 01:34 rpcutil.ddl
-rw-r--r-- 1 root root 3120 May 21 01:34 rpcutil.rb
1.1.2 载入Agent插件
[root@puppetserver rpms]# mco #客户端默认在自动载入
The Marionette Collective version 2.2.4
usage: /usr/bin/mco command <options>
Known commands:
completion facts find
help inventory ping
plugin puppet rpc
Type '/usr/bin/mco help' for a detailed list of commands and '/usr/bin/mco help command'
to get detailed help for a command
[root@agent1 ~]# /etc/rc.d/init.d/mcollective restart
Shutting down mcollective: [ OK ]
Starting mcollective: [ OK ]
1.1.3 验证Agent插件是否被载入
[root@puppetserver rpms]# mco inventory agent1.kisspuppet.com #查看节点agent1是否已经载入puppet插件
Inventory for agent1.kisspuppet.com:
Server Statistics:
Version: 2.2.4
Start Time: Thu Oct 03 16:09:03 +0800 2013
Config File: /etc/mcollective/server.cfg
Collectives: mcollective
Main Collective: mcollective
Process ID: 8902
Total Messages: 3
Messages Passed Filters: 3
Messages Filtered: 0
Expired Messages: 0
Replies Sent: 2
Total Processor Time: 0.46 seconds
System Time: 0.12 seconds
Agents:
discovery puppet rpcutil
Data Plugins:
agent fstat puppet #已经载入puppet插件
resource
Configuration Management Classes:
No classes applied
Facts:
mcollective => 1
1.1.4 从MCollective中运行Puppet
在运行命令之前，可以在节点查看puppet日志和puppetd服务的启停来判断命令是否调用了puppetd进程。
[root@puppetserver ~]# mco puppet --noop --verbose status #查看节点agent守护进程状态
Discovering hosts using the mc method for 2 second(s) .... 2
* [ ============================================================> ] 2 / 2
agent2.kisspuppet.com: Currently stopped; last completed run 9 hours 35 minutes 36 seconds ago
agent1.kisspuppet.com: Currently stopped; last completed run 9 hours 35 minutes 34 seconds ago
Summary of Applying:
false = 2
Summary of Daemon Running:
stopped = 2
Summary of Enabled:
enabled = 2
[root@puppetserver rpms]# mco puppet -v runonce
Discovering hosts using the mc method for 2 second(s) .... 2
* [ ============================================================> ] 2 / 2
agent1.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
agent2.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
---- rpc stats ----
Nodes: 2 / 2
Pass / Fail: 2 / 0
Start Time: Thu Oct 03 16:12:03 +0800 2013
Discovery Time: 2007.23ms
Agent Time: 3591.72ms
Total Time: 5598.94ms
备注：当使用MCollective运行Puppet时，要求在所有被管理的节点上Puppet agent守护进程都需要被关闭。在每次使用mco puppet -v runonce命令调用puppetd agent时，MCollective都会产生一个新的Puppet进程。这个进程会和任何已经运行的Puppet agent守护进程产生功能性的重复。
当Puppet使用--runonce参数运行时，agent会在后台运行。所以虽然MCollective成功运行了Puppet，但实际上的Puppet agent运行可能http://kisspuppet.com/2013/11/10/my-fact/并不成功。需要查看Puppet报告来确定每一个Puppet agent运行的结果。MCollective返回的OK值表示MCollective服务器成功地启动了puppetd进程并且没有得到任何输出。
1.2 安装facter插件（测试多次发现存在不稳定性）
注意：通过facter插件获取节点facter变量信息不是很稳定，因此可将节点facts信息通过inline_template写入/etc/mcollective/facts.yaml中，并在/etc/mcollective/server.cfg中设置factsource = yaml，这样MCollective客户端只需要每次读取这个文件中的facter变量即可。而且在本地目录/var/lib/puppet/yaml/facts/也会生成一份节点的facter信息，模块部分信息如下：
class mcollective::facter {
file{"/etc/mcollective/facts.yaml":
owner => root,
group => root,
mode => 0440,
loglevel => debug, # reduce noise in Puppet reports
content => inline_template('<%= scope.to_hash.reject { |k,v| k.to_s =~ /(uptime.*|path|timestamp|free|.*password.*|.*psk.*|.*key)/ }.to_yaml %>'),
}
}
[root@agent1 ~]# yum install mcollective-facter-facts
[root@agent1 rpms]# ll /usr/libexec/mcollective/mcollective/facts/
total 12
-rw-r--r-- 1 root root 422 Feb 21 2013 facter_facts.ddl
-rw-r--r-- 1 root root 945 Feb 21 2013 facter_facts.rb
-rw-r--r-- 1 root root 1530 May 21 01:34 yaml_facts.rb
[root@agent1 ~]# vim /etc/mcollective/server.cfg
…
# Facts
#factsource = yaml #注释掉
factsource = facter
plugin.yaml = /etc/mcollective/facts.yaml
[root@agent1 rpms]# /etc/rc.d/init.d/mcollective restart
Shutting down mcollective: [ OK ]
Starting mcollective: [ OK ]
[root@puppetserver rpms]# mco inventory agent1.kisspuppet.com #查看节点agent1是否加载了facts插件
Inventory for agent1.kisspuppet.com:
Server Statistics:
Version: 2.2.4
Start Time: Thu Oct 03 16:31:47 +0800 2013
Config File: /etc/mcollective/server.cfg
Collectives: mcollective
Main Collective: mcollective
Process ID: 9485
Total Messages: 37
Messages Passed Filters: 33
Messages Filtered: 4
Expired Messages: 0
Replies Sent: 32
Total Processor Time: 0.74 seconds
System Time: 0.21 seconds
Agents:
discovery puppet rpcutil
Data Plugins:
agent fstat puppet
resource
Configuration Management Classes:
No classes applied
Facts: #可以看到获取的节点facter信息（获取信息需要一些等待时间）
architecture => x86_64
augeasversion => 0.10.0
bios_release_date => 07/02/2012
bios_vendor => Phoenix Technologies LTD
bios_version => 6.00
blockdevice_fd0_size => 4096
…
uptime_days => 0
uptime_hours => 20
uptime_seconds => 74506
uuid => 564DFBAB-CADC-FC69-36CA-955BFDB30F43
virtual => vmware
[root@puppetserver rpms]# mco facts lsbdistdescription -v #使用mco facts命令对操作系统类型进行显示
Discovering hosts using the mc method for 2 second(s) .... 2
Report for fact: lsbdistdescription
Red Hat Enterprise Linux Server release 5.7 (Tikanga)found 1 times
agent2.kisspuppet.com
Red Hat Enterprise Linux Server release 5.8 (Tikanga)found 1 times
agent1.kisspuppet.com
---- rpc stats ----
Nodes: 2 / 2
Pass / Fail: 2 / 0
Start Time: Thu Oct 03 16:59:04 +0800 2013
Discovery Time: 2004.83ms
Agent Time: 67.32ms
Total Time: 2072.15ms
[root@puppetserver rpms]# mco facts lsbdistdescription #使用mco facts命令对操作系统类型进行统计
Report for fact: lsbdistdescription
Red Hat Enterprise Linux Server release 5.7 (Tikanga)found 1 times
Red Hat Enterprise Linux Server release 5.8 (Tikanga)found 1 times
Finished processing 2 / 2 hosts in 79.15 ms
[root@puppetserver rpms]# mco facts -v --with-fact hostname='agent1' memoryfree #查看主机agent1的剩余内存
Discovering hosts using the mc method for 2 second(s) .... 1
Report for fact: memoryfree
795.13 MB found 1 times
agent1.kisspuppet.com
---- rpc stats ----
Nodes: 1 / 1
Pass / Fail: 1 / 0
Start Time: Thu Oct 03 17:02:13 +0800 2013
Discovery Time: 2005.65ms
Agent Time: 49.37ms
Total Time: 2055.03ms
1.3 使用元数据定位主机
1.3.1 使用默认facter元数据定位主机
1.3.1.1 触发所有系统为RedHat，版本为5.7的所有节点puppetd守护进程
[root@puppetserver rpms]# mco puppet -v runonce rpc --np -F operatingsystemrelease='5.7' -F operatingsystem='RedHat'
Discovering hosts using the mc method for 2 second(s) .... 1
agent2.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
---- rpc stats ----
Nodes: 1 / 1
Pass / Fail: 1 / 0
Start Time: Thu Oct 03 17:03:56 +0800 2013
Discovery Time: 2008.09ms
Agent Time: 1187.69ms
Total Time: 3195.78ms
1.3.1.2 触发所有系统为RedHat，kernel版本为2.6.18的所有节点puppetd守护进程
[root@puppetserver rpms]# mco puppet -v runonce rpc --np -F kernelversion='2.6.18' -F operatingsystem='RedHat'
Discovering hosts using the mc method for 2 second(s) .... 2
agent2.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
agent1.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
---- rpc stats ----
Nodes: 2 / 2
Pass / Fail: 2 / 0
Start Time: Thu Oct 03 17:06:15 +0800 2013
Discovery Time: 2004.32ms
Agent Time: 1308.34ms
Total Time: 3312.66ms
1.3.2 使用自定义facter元数据定位主机
备注：使用自定义facter元数据可以更加灵活的定位主机，如何定义fact可参考博文《通过自定义fact增强MCollective推送更新元数据的灵活性》
1.3.2.1 在agent1上定义facter my_apply1和my_apply2
[root@agent1 mcollective]# facter -p | grep my_apply
my_apply1 => apache
my_apply2 => mysql
1.3.2.2 在agent2上定义facter my_apply2和my_apply3
[root@agent2 mcollective]# facter -p | grep my_apply
my_apply2 => mysql
my_apply3 => php
1.3.2.3 在MCollective客户端测试节点自定义facter是否正确
[root@puppetserver facter]# mco inventory agent1.kisspuppet.com | grep my_apply
my_apply1 => apache
my_apply2 => mysql
[root@puppetserver facter]# mco inventory agent2.kisspuppet.com | grep my_apply
my_apply2 => mysql
my_apply3 => php
1.3.2.4 通过自定义facter定位主机触发更新
[root@puppetserver facter]# mco puppet -v runonce mco facts -v --with-fact my_apply3='php' #筛选节点facter变量my_apply3=php的主机进行触发puppetd守护进程
Discovering hosts using the mc method for 2 second(s) .... 1
* [ ============================================================> ] 1 / 1
agent2.kisspuppet.com : OK
{:summary=> "Started a background Puppet run using the 'puppet agent --onetime --daemonize --color=false --splay --splaylimit 30' command"}
---- rpc stats ----
Nodes: 1 / 1
Pass / Fail: 1 / 0
Start Time: Thu Oct 03 23:33:54 +0800 2013
Discovery Time: 2005.35ms
Agent Time: 1078.86ms
Total Time: 3084.21ms