basework for #50

The main motivation behind doing this right now is to help customers who
need to know precisely what command(s) Bolt will run as sudo. The task
can be made into a predictable, constant invocation.

In RHEL, VERSION_ID can be something like "7.3", rather than just "7".
So for EL os's strip out the minor version when calculating the platform
tag.

This is not a full overhaul of the docs, but it at least clarifies the
basic introductory text of the README file.

Just a stub for now

Otherwise, errant agent runs could interfere with other aspects of the
upgrade process

To consolidate, and to better reflect current state of module

Node manager fix

Doing what the linter and Rubocop told me to

This commit allows you to manage the license key file
associated with PE.  This file is managed in the same
way the r10k private key file is currently managed in
this tool per the request of @reidmv

* When copying private key and license, flatten and compact target
  list - Bolt doesn't play well with undef as a target
* Because the file/content pattern is being used twice now, relegate
  details to a function (DRY)

Manage license file

We have observed setting r10k_private_key by default can cause conflicts
with advanced code manager configuration keys passed in by pe_conf_data.
To avoid the need to zero-out the r10k_private_key value, if the
top-level r10k_private_key settings are not given to peadm::provision,
do not set a pe.conf value for r10k_private_key.

Eliminate default private key parameter value

Update function to accept either string or undef data type.

…present without a standalone puppetdb replica.

To try and avoid dealing with the complexity of PuppetDB database host
picking at the end of the manifest, see if having a sane class parameter
default solves the problem cleanly. If it does, we can do a simple
switch on replica/no-replica to create the B groups.

It doesn't like having an optional parameter come before a non-optional
one. Since parameters are moving around a bit, add some comments to
clarify their intended usage.

Node manager switch

…s when installing an agent

updated the agent_install.sh file sed statement for installation flag…

This is because the curl|bash installer overrides certname and server
settings when installing. In the event a server needs to use a certname
which is NOT the same as the system's configured hostname, the certname
must be set manually at agent install time.

DRY up the code a bit

The only way to set the certname right now for hosts configured with the
PE installer is to pre-stage a puppet.conf file with the certname value
set. This commit does that for PE installer hosts in order to allow all
infrastructure nodes to have their certnames set for them, based on the
Bolt target name.

If the target name is different from the IP address
used to reach the host, then an inventory file should be used.

This commit removes the hard check that target names be equal to
hostnames. This commit is in support of other changes added to allow the
use of certnames which are not the hostname.

Set certnames to target names

Include some notes on how to use an inventory file to specify a target
certname that is different from the uri used to access the target.

The key "nodes" was deprecated in favor of "targets" in later versions
of Bolt 1.x, and obsoleted/replaced in Bolt 2.x. Updating docs to be
current, valid inventory file sytnax.

The peadm::pe_install task as written will probably never technically
fail, even when the puppet-enterprise-installer script exits non-zero.

This commit updates the task to exit with whatever exit code the
puppet-enterprise-installer script exited with. Not a perfect fix but a
step in the right direction.

Fixes #71

Fail peadm::pe_install task when installer fails

It is no longer required that the hostname of a system match the target
name for the system given to Bolt. This commit is an additional update
to the docs to reflect that.

Previously, when upgrading a standard architecture deployment, after the
master runs the puppet-enterprise-installer script auth for compilers
needs to be restored by a puppet agent run. However, the upgrade plan
only performed a run on the dedicated database host, which doesn't exist
in a standard install (the databases are all on the master).

This commit reworks the upgrade sequence such that the master and
replica recieve puppet runs at the appropriate points, along with the
database targets, to ensure that regardless of which node the
database(s) are located on the node will have a Puppet run to finalize
its config before upgrade moves on to the compilers.

Fixes #73

Rework upgrade sequence to support Standard arch

Previously, peadm used pp_application and pp_cluster for this purpose.
This conflicted with some existing customer's usage of those attributes.
In order to avoid collision with customer OID usage, this commit
modifies peadm to use custom OIDs.

In the event a csr_attributes.yaml file is already present, don't
overwrite it; instead, merge in the values we need to any values already
present.