- Nextcloud + OnlyOffice
- *arr media management series (Lidarr, Sonarr, etc)
- Gitea
- Vaultwarden
- PiHole
- Jellyfin
- Wiki-js
- Lemmy
- Prometheus/Grafana/Loki
Currently all containerised running on a debian VM on a Rockylinux Qemu/KVM hypervisor. Initially I was using rocky+podman but inevitably hit something I wanted to run that just straight up needed docker and was too much effort to try and get working. 🤷
Hardware is an circa 2012 gaming machine with a few ZFS raids for all of my Linux ISOs. It lives an extremely tortured existence and longs for the sweet release of death.
Toying with the idea of migrating it all to on-prem virtualised kubernetes cluster using helm charts to manage the stacks and using NFS mounts for persistent storage because I hate myself (and to upskill I guess)
What about you?
Everything is running in containers on an Unraid server
With disks at idle/spun down, it consumes roughly 25W.
I have a very similar setup minus the iot and metric related services. I’m managing the services with Docker Compose on unRAID.
What’s the reasoning behind using docker compose on unraid, instead of the built in docker implementation?
For a couple reasons
Store and version configs in git. I realize unRAID provides flash drive backup (using git also), but this allows me to spin up my setup on another machine that may not be running unRAID. Helped recently when I switched away from Proxmox.
Allows me to group services with their dependencies. ( e.g. postgres, redis, etc ) Also can help isolate service groups from each other. Avoiding port conflicts on common db ports for example. Downside being may have more than one database, redis, etc.
Note, there is an unRAID docker compose plugin so you can still get easy access management buttons to start, stop, view logs, and edit services.
Personally I use it for a couple services that would be difficult to run separately (ie: deemix + lidarr). I’m also planning on moving all of my services with databases over to compose. I do lose a couple other QOL features but I still prefer this approach to start/stop all related containers instead of manually having to close each one.
Can you elaborate on your host?
What exactly?