Shaken Fist is a deliberately minimal cloud. You can read more about Shaken Fist at https://github.com/shakenfist/shakenfist and shakenfist.com -- this repository is the deployment and CI tooling for the project, and therefore not a great place to start your journey.

Build an acceptable deployment, noting that only Ubuntu is supported. Deployments are based on Hashicorp terraform configuration found in deploy/ansible/terraform/ where "cloud" is one of the following at the time of writing:

• aws: Amazon EC2 bare metal (3 node cluster)
• aws-single-node: Amazon EC2 bare metal (1 node cluster), note that the CI tests will not pass on a single node cluster
• gcp: Google cloud, using nested virtualization (see additional steps below). This deployment option is not for production use because of its lack of bare metal support and low MTUs affecting virtual network performance. We mainly use it for CI testing.
• metal: Baremetal provisioned outside of terraform
• openstack: OpenStack, using nested virtualization
• shakenfist: Shaken Fist can self host, this is useful for CI for example

sudo apt-get install ansible tox pwgen build-essential python3-dev python3-wheel curl
git clone https://github.com/shakenfist/deploy
ansible-galaxy install andrewrothstein.etcd-cluster andrewrothstein.terraform andrewrothstein.go

On Google Cloud, you need to enable nested virt first:

# Create an image with nested virt enabled (only once)
gcloud compute disks create sf-source-disk --image-project ubuntu-os-cloud \
    --image-family ubuntu-1804-lts --zone us-central1-b
gcloud compute images create sf-image \
  --source-disk sf-source-disk --source-disk-zone us-central1-b \
  --licenses "https://compute.googleapis.com/compute/v1/projects/vm-options/global/licenses/enable-vmx"

The "metal" installation option can be used to create a test cluster on VMWare ESXi hypervisors.

Virtual machines hosted under ESXi need two CPU options enabled.

Hardware virtualization:
    Expose hardware assisted virtualization to the guest OS

Performance counters:
    Enable virtualized CPU performance counters

The installer will also enforce the following sanity checks:

• That KVM will operate on your machines. This is generally fine unless you're using virtual machines at which point nested virtualization needs to be enabled.
• That your network interface MTU is greater than 2,000 bytes. This is required because the VXLAN mesh our virtual networks use add overhead to packets and a standard MTU of 1500 bytes for the physical network will result in packets being fragmented too frequently on the virtual networks. You can set a higher MTU if you desire, I generally select 9,000 bytes.

Each deployment takes slight different arguments. I tend to write a small shell script to wrap these up for convenience. Here's an example for a baremetal deployment:

$ cat ~/sf-deploys/cbr-remote.sh 
#!/bin/bash

export CLOUD=metal
export ADMIN_PASSWORD="p4ssw0rd"
export FLOATING_IP_BLOCK="192.168.20.0/24"
export BOOTDELAY=0

# Tests are not currently safe for clusters whose data you are fond of
export SKIP_SF_TESTS=1

./deployandtest.sh

I then execute that script from the deploy/ansible directory to do a deployment or upgrade. Note that running the CI suite will destroy the contents of your cloud so be sure to use SKIP_SF_TESTS if you're upgrading a cluster with real users.