Introduction

Lilypad is put together using the float configuration management toolkit for container-based services. It is a series of Ansible plugins and roles glued together to provide a simple container-oriented environment. This can be rolled into your own Ansible configuration, or used separately.

Monitoring, alerting, log-collection and analysis, DNS and Let's Encrypt certificates for all of the services included are handled automatically. Please see the float reference documentation for further details.

Pre-requisites

You need at least three different remote machines. These can be bare-metal, or virtual machines (eg. KVM). They should have a minimal Debian Buster installation and be reachable by SSH.

One machine will act as a reverse proxy, and provide the infrastructure front-end. The second machine will run as an application server that the reverse proxy talks to, it runs the LEAP web API, its gateway selection service, and the infrastructure that provides monitoring and alerting. The third+ machine(s) are the VPN gateways, they *** will need two publicly addressable IP addresses ***.

The machines should be considered to be fully managed by this framework when things have been deployed. It will modify the system-level configuration, install packages, start services, etc. However, it assumes that certain functionality is present, either managed manually or with some external mechanisms: network configuration, partitions, file systems, and logical volumes must be externally (or manually) managed. SSH access and configuration must be externally managed unless you explicitly set enable_ssh=true (and add SSH keys to your admin users), in which case deployment will take over the SSH configuration.

You will need to pick a subdomain and delegate its DNS for the system to manage. For example, if your domain is example.com, then you could delegate, for example, the subdomain float.example.com. You would do this by adding a NS record for float.example.com that points to ns1.example.com and then an A record for ns1.example.com that points to the IP address you use for the reverse proxy host (note: not the gateway IP).

The following commands should be run locally on your computer in order to install and deploy Lilypad on the remote machines.

0. Clone the float repository

...and enter it

git clone https://0xacab.org/leap/container-platform/lilypad
cd lilypad

1. Install the float and LEAP platform pre-requisites

This installation guide is tested on Debian Bullseye. Other Linux distributions might need additional steps to install all requirements in the correct version.

sudo apt-get install golang build-essential bind9utils git

go install git.autistici.org/ale/x509ca@latest
go install git.autistici.org/ale/ed25519gen@latest
go install git.autistici.org/ai3/go-common/cmd/pwtool@latest
export PATH=$PATH:$HOME/go/bin

We'll use virtualenv to manage and install python packages:

virtualenv -p /usr/bin/python3 venv
source ./venv/bin/activate
pip install -r ./requirements.txt

This will create a virtual environment where we can install+version control specific python dependencies.

When working on this project, if you're in a new shell, you'll need to run the source ./venv/bin/activate command again to re-enter the virtual environment.

2. Initialize the ansible vault

... by creating a password file. Keep the public user ID of your OpenPGP keys at hand:

tr -dc 'A-Za-z0-9' < /dev/urandom | head -c 26 | gpg -ea -o .ansible_vault_pw.gpg

The resulting .ansible_vault_pw.gpg will be automatically decrypted by Ansible at runtime (use of an agent, such as gpg-agent is advised).

Configure your local environment to know where the ansible vault password is located:

export ANSIBLE_VAULT_PASSWORD_FILE=.ansible_vault_pw.gpg

This environment variable will only be set for this shell, you will need to add it to your shell environment initialization file so it will be set automatically everytime.

3. Customize the environment

Open hosts.yml and change floatapp1 to your app host's hostname, and specify the ansible_host and ip values to be the IP addresses for that host. If you have more than one app server, then you would just create a copy of this block, modifying the values, being sure to keep the 'backend' group assigned to each one.

Configure the front-end reverse proxy with in the same way, change the floatrp1 hostname to your hostname, and the ansible_host and ip to the IP it should have, and set the location value to where this server is located. For the egress_ip, put the secondary gateway ip.

Then edit group_vars/all/config.yml and set your domain_public to the subdomain name that you delegated (eg. float.example.com), the domain can be set to infra.example.com as this is the internally managed domain.

The config.yml contains a list of admins, a default hashed password and a set of ssh keys that will be able to connect to the system as root. If you do not change this password, then the user 'admin' and password 'password' are used. To change the hashed password you can run

pwtool <type-here-your-password>

and paste the output into the password variable. Have a look at the common operators playbook for additional options, such as setting up OTP or U2F tokens.

This config.yml also contains the credentials for an updated geoip database. The geoip_account_id and geoip_license_key values must be changed, you can register for an account on maxmind.com to obtain these. The geoip service helps end users to choose a gateway near them (usually faster).

Then edit group_vars/all/gateway_locations.yml, group_vars/all/provider_config.yml to match your environment.

NOTE: The value of location for a VPN gateway host, and the location keys in gateway_locations.yml, must match exactly one of these city names.

4. Generate credentials

... by running the init-credentials playbook. This will ansible-vault-encrypt the resulting secret files under credentials/. Note: this is not the built-in float init-credentials, rather this is the LEAP provided one, which will instantiate the float init-credentials when it is finished.

float/float run playbooks/init-credentials

You should not see any red text in this process, if you do, stop now.

This will generate service-level credentials, which are automatically managed by the toolkit and are encrypted with ansible-vault. These include the internal X509 PKI for TLS service authentication, a SSH PKI for hosts, and application credentials.

5. Consider comitting the generated credentials

... to git, and pushing them to a repository. All auto-generated credentials are stored in the credentials_dir - you will want to ensure that these are properly encrypted, checked into a git repository and kept private. The secret material is encrypted with ansible-vault, so it cannot be read without the access to the .ansible_vault_pw. If you commit these files, and push them to a respository, then you can share them with other admins, but be aware that these are secrets that should not be shared with anyone but trusted admins. If you gpg encrypted the .ansible_vault_pw, then that file is also encrypted and could also be committed.

6. Ensure SSH access

Lilypad uses elliptic curves for ssh, ed25519. Make sure you can ssh to the hosts as root without being prompted for a password every time after having verified and accepted the correct host key. Try to login:

ssh -i ~/.ssh/id_ed25519 root@float.example.com

7. Deploy the configuration

Run:

float/float run site.yml 

This will take some time to finish, as it needs to download packages and Docker images and configure everything.

8. Update servers

Run:

float/float run float/playbooks/apt-upgrade.yml

Congratulations. You have successfully installed and deployed the LEAP platform! You should read the documentation about how to perform common operations.

Testing

Certificate authority from provider: leap.ca

Make a CSR/key

sign cert against CA

make sure the x509 v3 extensions exist: x509.ExtKeyUsageClientAuth x509.KeyUsageDigitalSignature

/usr/sbin/openvpn --client --remote-cert-tls server --tls-client --remote 37.218.241.84 80 --proto tcp --verb 3 --auth SHA1 --keepalive 10 30 --tls-version-min 1.2 --dev tun --tun-ipv6 --ca ./ca.pem --cert ./testopenvpn.crt --key ./testopenvpn.key

Reference: https://0xacab.org/leap/vpnweb/blob/master/certs.go#L37

    ExtKeyUsage: []x509.ExtKeyUsage{x509.ExtKeyUsageClientAuth},
    KeyUsage:    x509.KeyUsageDigitalSignature,
    CommonName: UNLIMITED
    subjectkeyID: random
    serial: random

Integration Testing

Integration tests can be run to: * check that public endpoints for built-in services are reachable * check that no Prometheus alerts are firing

These tests can be run from your Ansible directory using the float command-line tool:

/path/to/float/float run integration-test

The test suite requires a small amount of configuration in order to run on a non-test environment, as it needs admin credentials in order to automatically test SSO-protected services. This is stored in a YAML file, you can point the test suite at your own test parameters using the TEST_PARAMS environment variable, e.g.:

env TEST_PARAMS=my-params.yml /path/to/float/float run integration-test

The built-in test parameters configuration uses the credentials for the default admin user used in test environments (admin/password):

---
priv_user:
  name: admin
  password: password

Testing float

    apt install qemu-kvm libvirt-clients libvirt-daemon-system bridge-utils vagrant vagrant-libvirt
    adduser micah libvirt
    adduser micah libvirt-quemu
    float create-env --vagrant --num-hosts 2 test
    cd test; vagrant up

FAQ

Why is there a '[openvpn]' group, but no host attached to it?

You might have noticed that site.yml has a hosts parameter with roles assigned to them, and the actual hosts defined in site.yml are connected to the hosts.yml groups parameter. The hosts.yml has floatrp1 with the groups: [frontend], but there is no host which has the [openvpn] group attached to it.

For the 'openvpn' service, there is a scheduling_group, which sets the scope of the possible hosts that the service will be scheduled onto. Float will create automatically a 'openvpn' group, containing just the hosts that 'openvpn' is running on. We did not define an 'openvpn' group in the hosts.yml ansible inventory, yet such a group is automatically created by float, and you can use it in Ansible. This 'openvpn' group is a subset of the scheduling_group.

"where can I run openvpn" -> scheduling_group (frontend)

"where is openvpn actually running"* -> "openvpn" group