Coffee & Beer

Rantings and Ravings of the technical sort

Getting Kvm Domain Info Into Puppet Facts

| Comments

While I’m yet to find the holy grail of getting a KVM host to be aware of which ‘dom0’ its being hosted on, this bit of ruby for a custom puppet facts is a step there, the other way around:

(vms-running.rb) download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

require 'facter'
begin
        Facter.compute_node
rescue
        Facter.loadfacts()
end


unless  `rpm -qa | grep ruby-libvirt`.empty?
        if (Facter.value("puppet_class_kvm_host") == "true") then
                Facter.add("kvm_vms") do
                        setcode do
                                require 'libvirt'
                                conn = Libvirt::open('qemu:///system')
                                if (conn.num_of_domains == 0 ) then
                                        kvm_vms = "NO VMS"
                                        conn.close
                                        kvm_vms
                                else
                                        vm_doms = Array.new
                                        conn.list_domains.each do |domid|
                                                dom = conn.lookup_domain_by_id(domid)
                                                vm_doms.push(dom.name)
                                        end
                                        vm_doms=vm_doms.join(" ")
                                        conn.close
                                        vm_doms
                                end
                        end
                end
        end
end

Which will return a nice fact like:

kvm_vms => Host1 Host2 Host 3

or if none are running:

kvm_vms => NO VMS

its ugly, but it works!

Finding Compute Resources With Puppet Facts

| Comments

One question we get asked at wrk a good bit centers around finding compute resources that match a specific set of specifications.Since the compute system we haev are of all sort of various ages and hardware types, jobs that need 8GB of memory might not work everywhere, but most places.Same goes for a specific number of cores (or, minimal #), per node. Some have 4, some have 12+.

Well, facter, and therefore puppet, knows about these things. And there is a REST api we can query. And I needed an excuse to do a little ruby:

(find_compute.rb) download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69

#!/usr/bin/env ruby

require 'yaml'
require 'puppet'

puts "Welcome to the Compute Finder script!"
puts "This script aims to help you locate compute nodes based on simple requirements"
puts "Such as the minimum amount of RAM or # of processor cores"
puts "---"
puts "---"
puts "What minimum amount of ram would you like?"
puts "(in GB, leave blank for no minimum)"
mem=gets.chomp
puts "What minimum number of processor cores would you like?"
puts "(leave blank for no minimum)"
procs=gets.chomp

sleep 1

search ="facts.compute_node=true"
if mem != ""
  search << "&facts.memorysize.ge=#{mem}"
end

if procs !=""
  search << "&facts.processorcount.ge=#{procs}"
end

if (procs == "" and mem == "" )
  puts "You must specify some requirments!"
  exit
end
puppet_base = "https://puppet:8140"
path="/production/facts_search/search?"


puts "Finding matching nodes, please wait..."
cmd = "curl -s -k -H \"Accept: yaml\" \"#{puppet_base}#{path}#{search}\""
ans = %x[#{cmd}]
nodes = YAML::load(ans)
nodearr = []
nodes.sort.each do |node|
  nodearr.push(node)
end
puts nodearr
puts "Would you like more info on the nodes found? This may take bit longer..."
puts "(y/n)"
more=gets.chomp
if more == "y"

  node_array = []
  nodes.each do |node|
      node_url="https://puppet:8140/production/facts/#{node}"
      node_cmd="curl -s -k -H \"Accept: yaml\" #{node_url}"
      node_ans= %x[#{node_cmd}]
      node_array << YAML::load(node_ans)
  end

  results =[]
  node_array.each do |node|
      name = node.name
      facts = node.values
      ram=facts['memorysize']
      procs=facts['processorcount']
      results.push("Host: #{name} RAM: #{ram} ProcessorCores: #{procs}")
  end
  results.sort!
  puts results
end

The output of which looks like:

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

Welcome to the Compute Finder script!
This script aims to help you locate compute nodes based on simple requirements
Such as the minimum amount of RAM or # of processor cores
---
---
What minimum amount of ram would you like?
(in GB, leave blank for no minimum)
64
What minimum number of processor cores would you like?
(leave blank for no minimum)
12
Finding matching nodes, please wait...
node01.local
node02.local
node03.local
node04.local
Would you like more info on the nodes found? This may take bit longer...
(y/n)
y
Host: node01.local RAM: 94.28 GB ProcessorCores: 12
Host: node02.local RAM: 94.28 GB ProcessorCores: 12
Host: node03.local RAM: 86.39 GB ProcessorCores: 12
Host: node04.local RAM: 251.89 GB ProcessorCores: 32

Puppet Facts About Puppet Classes

| Comments

So, obviously,we use puppet a great deal. One thing a few of us have wanted for a while was the ability to ssee that nodes/node groups would be affected by a change to a particular class. In other words, what hosts get $This class?

Now, nodes are aware what classes they get, but the puppet master seemingly is not, not really. It compiles teh catalog on a per-run basis, and while durring that it DOES know what classes a system gets, it doesn’t really store that info. So, a simple custom fact to pull that data in from each host:

(puppet_classes.rb) download
1
2
3
4
5
6
7
8
9
10

#list puppet classes, on per fact
require 'facter'

IO.popen('cat /var/lib/puppet/classes.txt | tail -n+3').readlines.each do |line|
        Facter.add("puppet_class_#{line}") do
                setcode do
                        "true"
                end
        end
end

So, now we get, for each host, a bunch of facts like:

1

puppet_class_syslog => true

Reported from facter -p, and therefor sent back to teh puppet master, into a DB, which we can query directly, via teh puppet api, or via Foreman.