Keiran Sweet presented a case study on how his organization used Puppet to bring order to chaos in their environment. Originally, their Linux servers were inconsistently configured by hand and suffered from instability. They implemented Puppet for configuration management and the Foreman tool for provisioning. This centralized configuration of servers, provided self-healing, improved reporting, and integrated with their provisioning workflow. As a result, costs went down and stability and speed of deployments increased. Their fleet of servers could now be easily managed and had a consistent, automated configuration.

1. Keiran Sweet
Bringing order to chaos with Puppet

2. Bringing order
to Chaos with
Puppet
Keiran Sweet

3. About me
• UNIX Administrator in the Financial sector
• ~10 years UNIX experience across a
number of industries
• Using Puppet for 3+ years
• Devops advocate
• Likes automating things
• Likes keeping things simple

4. Obligatory..
disclaimer { keiran:
opinions => mine,
representing => myself,
}

5. Puppet ?
• What is it ?
• What can it do ?
• Who uses it ?
• What is it's role in Devops ?
• Similar to ?What is i

6. Example DSL #1
# Managing a User account
user { 'Keiran.Sweet':
ensure => 'present',
comment => 'Keiran Sweet',
gid => '10042',
groups => ['wheel'],
home => '/home/Keiran.Sweet',
shell => '/bin/bash',
uid => '10042',
}

7. Example DSL #2
# Deploying and managing a file
file { '/etc/security/limits.conf':
ensure => 'file',
source => 'puppet:///modules/ostune/limits.conf',
owner => 'root',
group => 'root',
mode => '644',
}

8. Case study

9. Case Study - Original Environment
• OS - Enterprise Linux
o RHEL 4 & 5
o CentOS 4 & 5
o RHEL 6 was on the way
• "Management" tools
o Satellite Server - RHEL Management
o Spacewalk Server - CentOS Management
o Glued together with Cobbler, DHCP, TFTP & PXE
o Numerous "rogue" servers
o External 3rd Party repositories
o Lots of hand crafted servers

10. Case Study - Original Environment
• Provisioning
o Kickstart via Satellite & Spacewalk
 Basic installs of the OS only
 Servers were built as localhost.localdomain
 Configuration done by hand post kickstart
 Packages weren't being utilised (tar, zips, scripts)
 VMTools, HP PSP's, HW Agents - Often missing
o DVD/CD Installs still being done
 Left disconnected from Satellite / Spacewalk
 Non-standard configurations
 Some talked to unknown repos on the internet (!)

11. Arriving in the office each morning...

12. Linux Team
Reaction

13. Case Study - Original Environment
• Result - Utter carnage
o Business was suffering
 Application & Server instability == Outages
 Delays in deployment of new platforms
 Delays in upgrading and deploying our
applications
o The Linux admins were suffering
 Flying blind - Unknown fleet composition
 Every issue resulted in a journey of discovery
 Inconsistent configurations made delivering other
infrastructure projects difficult - ie, Monitoring
 Wasn't enjoyable (not enough time for reddit)

14. Ain't nobody got
time fo dat

15. Case Study - What did we need ?
• Configuration Management
o Self healing
o Rich Reporting
o Vendor neutral
o Flexible
o Version control
o Integrates with provisioning workflow
• Software Management and Distribution
o Centralised, Simple & Scalable
o Secure (Signed Packages & TLS)
o Redundancy - Supported mirrors/slaves

16. Case Study - What we chose
• Puppet
o Configuration Management
o Self healing
o Rich reporting
o Cross platform
o Extensible
• Foreman
o Puppet node classifier (ENC)
o DHCP/TFTP/PXE Integration
o Puppet report visualisation
o Host inventory & CMDB

17. Case Study - How it was assembled

18. Case Study - Fixing Provisioning
Hostname
OS Version
MAC Address
Partition Table
Data Center

19. Case Study - Importing the fleet
• Install Puppet agent
• Boot strap the client
• Apply the standard
modules
• Test
Result:

20. Case Study - End result - Stability
• Costs are down
• Business applications are more stable
• React to requests more rapidly
• Users have a common experience
• Fleet management is much easier

21. Questions ?
• Contact
o Twitter: @keiran_s
o Email: Keiran@gmail.com
• Images
o Devops Reactions
o Reddit

22. Case Study - End result - Dashboard

23. Case Study - End result - Reports

24. Case Study - End result - Node data