Multiple node support with Swarm

[jaydrogers]

👉 Describe the problem

• Some users need to scale beyond a single server for better uptime

👥 Problem evidence & reach

• https://discord.com/channels/910287105714954251/1205536000210636852
• https://discord.com/channels/910287105714954251/1318249792824152166

🥰 Describe the "impact" on users?

• It would help them know they can achieve better uptime with Spin

🏆 How to solve this problem

Spin Ansible Collection

https://github.com/serversideup/ansible-collection-spin

• Expand upon the Ansible inventory to support multiple managers and workers

💯 How do we validate the problem is solved?

• Users can expand their .spin.yml to build their clusters
• Everything is clearly documented how to add managers and nodes
• Users can configure how to set certain services to managers and workers

[harrisonratcliffe]

I have this working but its a manual process, pretty straightfoward though.

Add your servers to the .spin.yml file and set the environment as production_workers for the workers instead of production.
Provision your servers as normal with spin provision
Join the nodes to your swarm cluster/manager - you can manually promote any workers to managers for high availability.
Deploy! 🚀

Note: I use Hetzner so I create a private network and add all my nodes to that. Then I modify the firewall that Spin creates (Pro feature) to allow TCP and UDP access on that private subnet for the swarm cluster to talk to eachother and set up a load balancer for tcp 80 -> 80 and 443 -> 443.

I have not tested this with spin deploy but I believe when I initally tried playing around with a cluster and Spin it failed to deploy so I use solely Github Actions.

I'm also working on a fork of serversideup/github-action-docker-swarm-deploy to allow multiple managers so, for example, if you had manager-1, manager-2 and manager-3, it will keep trying those managers if one is down. I'll share that and submit a PR once I have it working.