Ansible Automation Platform 2.x automated demo

What is this?

This repository holds some kcli plans and Ansible playbooks to automatically build a multi-node Ansible Automation Platform 2.x infrastructure on libvirt. Perfect for demoing and playing around.

Only thing that you need to do is deploy the appropriate plan on your laptop / workstation.

Pre-reqs

Before you can deploy the plan, you need to do a couple of things:

• clone this repo onto your local machine
• install kcli, see https://copr.fedorainfracloud.org/coprs/karmab/kcli/
• download rhel-8.4-x86_64-kvm.qcow2 from access.redhat.com and drop it into /var/lib/libvirt/images.
• create appropriate ~/.kcli/config.yml and ~/.kcli/profiles.yml files. Find some examples in the samples directory of the kcli GitHub repo.
• go to access.redhat.com, create a subscription allocation and download it. Place the manifest.zip file in the files directory of this repo.
• download ansible-automation-platform-setup-2.0.1-1-early-access.tar.gz from access.redhat.com and drop it into the files directory of this repo. The download location will change after AAP 2.1 is released properly.
• create a vault to store your credentials: rh_username and rh_password for your access.redhat.com credentials, and aap_admin_pw, aap_db_pw, hub_admin_pw and hub_db_pw for the application and database passwords. Easiest is to create it as group_vars/all/vault.yml.

Note: the default plan uses the same storage pool ('default') for all 6 VMs. That is a lot to stomach for most storage devices. If you have more than one disk, consider creating a storage pool on it and spreading your VMs over multiple pools.

Deploying the plan

What are the different available plans?

Because not everyone has a really beefy machine, I have created various plans to enable as many people as possible to enjoy automated demo deployment:

• aap2_cluster.yml deploys a three node controller cluster with 8192MiB of RAM each, an HAProxy load balancer with 2048MiB of RAM, an Automation Hub machine with another 8192MiB of RAM and a database - shared between the controller and Hub machines - with 4096MiB of RAM
• aap2_cluster_small deploys two controller nodes only, and brings down the amount of required memory significantly, by using only 4096MiB of RAM for the controller nodes, and only 2048MiB for the database, load balancer and Hub machines. This also requires passing required_ram=1800 to the installer, resulting in a cluster that is unsupported (but will run a simple demo just fine in my experience).
• aap2_cluster_only does the same as aap2_cluster_small, but does not deploy the Hub
• aap2_single_node does the same as aap2_cluster_only, but with a single node controller "cluster"

Actually running the plan

After filling in the pre-requisites - which you have to do only once per machine - you can deploy the plan from the root of this project by running:

kcli create plan aap-cluster -f plans/aap2_cluster_full.yml

To prevent the machines from updating and installing git, mc and vim, run this instead:

kcli create plan aap-cluster -f plans/aap2_cluster_plans.yml -P skip_updates=true -P skip_utilities=true

Wait for a couple of minutes for your plan to spin up and you are done :)

DNS

If you want, you can drop this content into /etc/NetworkManager/dispatcher.d/pre-up.d/97-lab.sh or something similar, and manually execute it the first time the redhat.lab network is created (after a reboot, this'll be automatic) to get DNS resolving for the redhat.lab domain:

#!/bin/sh
# This is a NetworkManager dispatcher script to configure split DNS for
# the 'crc' libvirt network.
# The corresponding crc bridge is recreated each time the system reboots, so
# it cannot be configured permanently through NetworkManager.
# Changing DNS settings with nmcli requires the connection to go down/up,
# so we directly make the change using resolvectl

export LC_ALL=C

if [ "$1" = lab ]; then
        resolvectl domain "$1" redhat.lab
        resolvectl dns "$1" 192.168.125.1
        resolvectl default-route "$1" false
        resolvectl dnssec "$1" no
fi

exit 0

In action

Here's a screencast of the script in action.

Pull requests, feature requests, etc.

Please send them :)

One thing tho: I know it is possible to split the parameters for each plan into a _default.yml file. That, however, brings it's own problems with calling the ansible playbook, the location of the vault and templates and other things. Therefore, I am inclined not to split them for now.