Background

Note:  this is a companion article to the blog post titled: "[HOWTO] Orchestration Basics - Using a Chef recipe to deploy Centrify and join Active Directory";  these posts are relatively identical, the only difference is the tool.

In the field, we often get inquiries on deployment, governance or orchestration frameworks.  Common questions range from: how do I deploy the Centrify client?  or How do I establish an approval flow for the Access Control model that you facilitate? 

Sometimes we also hear these questions:  We use Ansible | Bladelogic | Chef | Puppet in on premise or with our IaaS (cloud) deployment, how would I go about deploying your product using those tools?

These are important questions, especially in the age of private/public clouds and elastic environments.  Some of you are adopting frameworks like OpenStack, Amazon AWS or Microsoft Azure that underneath use these toolsets.

At Centrify we have invested heavily through the years to make sure that the products "just work".  From a deployment perspective, we our messaging is consistent:

• We always provide the native platform package
• Our tools are all Kerberized  (ready for automation)
• We also provide other tools like install.sh and Deployment manager that abstract the underlying OS package manager
• The Centrify agent automatically maintains its configuration and the Kerberos configuration based on Active Directory, and it uses AD specs for Sites and Services, DC location, offline cache.
• The configuration parameters can be managed by any tool, plus native Active Directory Group Policy.

In other words, Centrify software is easier to use in these scenarios than what the typical IT Infrastructure lead may think.

Disclaimers and Acknowledgements

• This article provides a "quick basic configuration";  in a true deployment you have to account for high-availability, replication, security, package integrity, supported platforms, supported versions, change control, etc. 
• All names, logos and trademarks used in this articles correspond to their existing owners.
• We are not Puppet, Chef or Bladelogic subject-matter experts, hence the use of a stand-alone script or recipe. 
• Puppet Master server/slaves/node architecture is outside the scope of this post
• This would not be possible without my customers/prospects asking and the tutorials available online.  Some great resources:
  - Introduction to Puppet:  https://docs.puppetlabs.com/guides/introduction.html
  - Willams' blog post "How to install Puppet in Stand-Alone mode on CentOS7"
  - Kudos to MaestroDev for the wget Puppet extension: https://forge.puppetlabs.com/maestrodev/wget

What is required?

• An Active Directory domain
• A Centrify zone (optional - if using Centrify in licensed mode/privilege management)
• A Centrify zone and a Computer Role  (optional)
• An Active Directory service account for joins and removals, plus a keytab for the account and a usable krb5.conf file.  Read this article if you want to know how to create the service account and obtain the keytab.
• A RHEL-based system with enough storage for the Centrify RPM packages for each platform (or just for the subset you need to support).  This system has to be set up as a YUM repository, as described in the the original orchestration article.
• A second RHEL-derivative system to install Puppet (see below), this system must have DNS settings configured correctly.

Note:  Although it's possible for you to follow this and put together a working prototype, I strongly-encourage that you really explore the concepts  of DevOps/infrastructure as code.  The whole philosophy promotes constant improvement, this means that if you expect this to be a "set it and forget it" solution, I advise that you realign your expectations.

Finally, By no means what's outlined here is ready for production.  Check out the reading list below. 

Example Diagram

Implementation Steps

Planning

The goal of this lab is to be able to deploy the Centrify bits in a RedHat, CentOS, Scientific or Oracle system in a consistent way.  Also, if you know what you're doing, you can extend this to other OSs, and platforms as well.

The 'core' process without checking for major dependencies or issues is to:

1. Install Centrify DirectControl
2. Authenticate against Active Directory with an account with minimal rights
3. Join Active Directory

Building Blocks

• YUM Repository with Centrify RPMs (detailed instructions here)
• AD account + keytab (steps outlined in a previous post)
• Usable krb5.conf:  You can copy this file from any Centrified system; however, depending on where you're onboarding the system, you want to edit the file only with the DCs that are reachable to the new system.

Make the krb5.conf and service account keytab available to your infrastructure

In the original article, we piggy-backed on the Apache webserver as the transport for our repository.  Now we're going to create another folder for utilities (utils for short) and copy the keytab and krb5.conf file.

If you prepared this for the Chef article, you can skip to Puppet installation.

1. Create a folder under /var/www/html
   $ sudo mkdir /var/www/html/centrify/utils
2. Copy the RPMs to the folder.
   $ cd /path/to/files
   $ sudo mv krb5.conf  /var/www/html/centrify/utils
   $ sudo mv ad-joiner.keytab  /var/www/html/centrify/utils
3. Set the proper permissions in the folder
   chmod -R ugo+rX /var/www/html/centrify/utils
4. Verify that the files are accessible via the web server (you may have to check the firewall settings)
   

Install Puppet and the wget Extension

1. Add the Puppet repository (you must find the proper repository for your RHEL version, version 7 shown)
   $ sudo rpm -ivh http://yum.puppetlabs.com/puppetlabs-release-el-7.noarch.rpm
2. Install the bits
   $ yum install puppet
3. Make sure your system's DNS settings are up to date (yes, no "localsystem.localdomain")
   Can you ping your own system by name and FQDN?
   What is the output of the hostname command?
   Run this:

   facter | grep hostname
   facter | grep fqdn

   if the output isincorrect, you must edit the /etc/hosts and /etc/resolv.conf and speak to your DNS admin to set things straight
4. Logout and log back in to verify the ruby path
   $  sudo puppet module install maestrodev-wget
   You should be ready to get going.

Create and test your stand-alone Puppet Script

To recap, the sequence to automate Installation and joins is as follows:

1. Retrieve a usable krb5.conf file
2. Retrieve the keytab of a valid service account with minimum rights to join systems to the target AD OU, to the target zone (if using in zone mode) and if adding to a Computer role, with rights to add to the target AD groups.
3. Install the Centrify Package and use the kinit tool to obtain a TGT
4. Run adjoin with the proper options.
5. Perform cleanup

Here is the Non-idenpotent Puppet Script:

# This stand-alone recipe will install the Centrify Agent on RHEL derivatives, 
# joins Active Directory and places the system in a Computer Role
# Notes: This recipe is not idempotent (achieving this is up to you!)

include wget

# Variables for my environment (see blog post)
# domain is the most basic parameter to join Active Directory

$adname = "centrify.vms"  
# Notice that I could not use "domain" like in the Chef example.  
# That word seems to be reserved in Puppet

# In Zone Mode (licensed with UNIX identity and Access Control) the zone
# parameter corresponds is where the system will be placed.  Not needed
# if working in workstation or express mode.

$zone = "Global"

# OU is where your computer object will be placed in Active Directory
# your ad-joiner account should be able to join systems to this container

$ou = " ou=servers,ou=unix"

# A Computer role is one of the ways to group systems and define access 
# control.  A system may be a member of multiple computer roles.  
# E.g.  a LAMP system may be accessible by Web Admins, Developers and 
# DBAs with different access rights and privileges.

$crole = "App Servers"

# nodes are the managed system in Puppet;  in a true deployment you can
# apply to individual systems or collections of systems.  Since this is not
# idempotent, you must specify the fqdn.

node "your-system-fqdn" {

# Centrify's utilities are Kerberized, this means that they will use the current
# user's Kerberos TGT to attempt the transaction against AD.  However, in a 
# virgin system, there are no working krb5.conf files, therefore kinit won't know
# how to find a KDC to authenticate against.  This is why we need a krb5.conf 
# file from a working system (or that points to a reachable Domain Controller), 
# in the previous blog entry, we piggy-backed on an Apache Web server to 
# serve those files (engcen6).

wget::fetch {"download a working krb5.conf":
  source             => 'http://engcen6.centrify.vms/centrify/utils/krb5.conf',
  destination        => '/temp/krb5.conf',
  timeout            => 0,
  verbose            => true,
  nocheckcertificate => true,
  before => Exec["kinit"]
}

# The keytab corresponds to a service account that has the minimal rights, in 
# this case, the rights to write a computer object in the designated container 
# (ou), centrify zone and the AD group that contains the "App Servers"computer
# role needless to say, you need to treat this file with care and if possible, 
# remove when complete.

wget::fetch {"download the ad-joiner keytab file":
  source             => 'http://engcen6.centrify.vms/centrify/utils/ad-joiner.keytab',
  destination        => '/temp/ad-joiner.keytab',
  timeout            => 0,
  verbose            => true,
  nocheckcertificate => true,
  before => Exec["kinit"]
}

# We leverage Puppet to ensure the files are present. This will be used later
# to guarantee proper sequencing.

file {"ad-joiner.keytab":
  ensure => present,
  path   => '/temp/ad-joiner.keytab'
}

file {"krb5.conf":
  ensure => present,
  path   => '/temp/krb5.conf'
}

# In this command, we authenticate against AD with the keytab of our service 
# account.  Note that we are using the usable krb5.conf file so kinit can reach
# a KDC (domain controller).  The end-result is that root (or sudo) user will
# have a TGT and you don't need to put keys, hashes or passwords in your 
# script.  The before/subscribe and require Puppet directives guarantee proper 
# sequencing.

exec {"kinit":
  command =>  "/bin/env KRB5_CONFIG=/temp/krb5.conf /usr/share/centrifydc/kerberos/bin/kinit -kt /temp/ad-joiner.keytab ad-joiner",
  before => Exec["adjoin"],
  subscribe => [
        File["/temp/ad-joiner.keytab"],
        File["/temp/krb5.conf"],
  ],
  require => Package["CentrifyDC"],
}


# In a pre-requiste blog entry, I outlined how to create a YUM repository for 
# RHEL and derivatives.  This means that you need a yum or apt repo with 
# the Centrify packages.  Puppet will simply make sure the package is present
# notice the differences, I declared this after the previous directives, when
# the package is a pre-requisite.

package {"CentrifyDC":
    ensure => 'installed',

}

# Finally we run adjoin.  At this point we are using the variables from my 
# environment.  Although in doing so, we broke the 'idempotent principles, I'm 
# certain that Puppet experts can find ways to improve on this.  

exec { "adjoin":
  command => "/usr/sbin/adjoin -z $zone -c $ou -R \"$crole\" -V $adname",
  require => Package["CentrifyDC"]

}

# In the cleanup phase, we clear the TGT and delete the utility files
# Although the keytab provides very specific limited AD rights, always make
# a habit of cleaning-up.

exec {'kdestroy':
  command => '/bin/env KRB5_CONFIG=/tmp/krb5.conf /usr/share/centrifydc/kerberos/bin/kdestroy',
  require => Exec['adjoin'],
}

exec {"/bin/rm -f /temp/*":
  require => Exec['kdestroy'],
}

} # End of Script

Verify the recipe

$ sudo puppet apply centrify-build.pp
Notice: Compiled catalog for engcen7.centrify.vms in environment production in 1.77 seconds
 Info: Applying configuration version '1458157258'
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Package[CentrifyDC]/ensure: created
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/File[krb5.conf]/ensure: created
 Info: /Stage[main]/Main/Node[engcen7.centrify.vms]/File[krb5.conf]: Scheduling refresh of Exec[kinit]
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Wget::Fetch[download a working krb5.conf]/Exec[wget-download a working krb5.conf]/returns: executed successfully
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/File[ad-joiner.keytab]/ensure: created
 Info: /Stage[main]/Main/Node[engcen7.centrify.vms]/File[ad-joiner.keytab]: Scheduling refresh of Exec[kinit]
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Wget::Fetch[download the ad-joiner keytab file]/Exec[wget-download the ad-joiner keytab file]/returns: executed successfully
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Exec[kinit]/returns: executed successfully
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Exec[kinit]: Triggered 'refresh' from 2 events
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Exec[adjoin]/returns: executed successfully
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Exec[kdestroy]/returns: executed successfully
 Notice: /Stage[main]/Main/Node[engcen7.centrify.vms]/Exec[/bin/rm -f /temp/*]/returns: executed successfully
 Notice: Finished catalog run in 38.32 seconds

Verification Video

Adjustments

There are absolutely many improvements that can be made.   Here are a few that come to mind (in no specific order):

• Check for Perl:  CentrifyDC requires Perl 5.8 and up.
• Chek for DC connectivity:  You can potentially check for connectivity to KDCs prior to attempting to authenticate.
• Inspect the name of the system:  In keeping with AD Naming conventions, we should check for length and uniqueness in the forest prior to join.
• Check if the system is already joined or in the desired zone/forest.
• Inspect the IP/DNS configuration:  Maybe the TCP/IP and DNS config is not right.
• Perform a Dynamic DNS update using addns:  After join, you can use addns to get the system registered with Microsoft DNS
• Obtain Certificates with ADCert:  If the system is used for SSL communications, you can get the cert manually using adcert.
• Manage centrifydc.conf:  There are some basic parameters that may not be manageable via GPO (e.g. DMZ) that can be managed from there.
• New Information:  Centrify DirectControl can provide information to Puppetregarding the AD topology.
• Add dependent packages:  LDAP Proxy or Centrify DirectAudit/DirectSecure come to mind.
• Most importantly:  This requires a Puppet infrastructure.  Make it as idempotent as possible!