nixos-servers/docker_swarm
2024-03-28 21:10:15 +01:00
..
inventory remove bancomart VM from docker swarm 2024-03-13 22:41:02 +01:00
playbooks move radicale to kubernetes 2024-03-28 21:10:15 +01:00
roles move radicale to kubernetes 2024-03-28 21:10:15 +01:00
.envrc add devShell for ansible swarm 2024-03-02 14:46:10 +01:00
ansible.cfg remove legacy code 2024-02-08 23:53:02 +01:00
flake.lock add devShell for ansible swarm 2024-03-02 14:46:10 +01:00
flake.nix add devShell for ansible swarm 2024-03-02 14:46:10 +01:00
README.md remove unused ansible group variables 2024-02-10 23:37:47 +01:00

Docker Swarm

On each of our machines, we deploy a virtual machine that participates in a Docker Swarm. However, only one VM is a manager (maestro) while two are workers (bancomart and vpay). This lack of redundancy in the cluster is deliberate: in case all nodes are down (e.g. misconfiguration or power outage) manual action would need to be taken in order to restore the cluster. In case of only one manager node, the cluster is always able to restore itself automatically.

While the operating system of the VMs is managed by NixOS, cluster creation and the deployment of workloads is done through Ansible. In my opinion, Ansible is a perfect fit for environments that tend to change a lot (such as a container cluster).

Stacks

On top of the Docker Swarm, we host several services deployed as Docker Stacks:

Secret decryption

The Ansible playbooks assume you have the password to Ansible vault present at ~/.config/home/ansible-vault-secret.