FOR SELF-HOSTERS
A homelab you can still operate in a year.
One person running more services at home than they can remember the details of.
The failure mode of a homelab is not downtime. It is forgetting — which host runs what, why that container has that flag, and whether the backups have ever worked.
WHERE YOU ARE NOW
What forgetting looks like.
You cannot remember which of three machines is running the thing that broke, or which compose file is the live one.
You put off updating anything because the last time you did, something unrelated stopped working and it took a weekend to find out what.
There is a cron job writing dumps to a disk. You have never restored one. You are fairly confident it works.
WHAT CHANGES
What actually changes.
The inventory maintains itself
Discovery re-reads each host on a schedule, so the answer to “what is running where” is current rather than remembered.
how it works →Snapshots before risky things
If you run Proxmox or libvirt, the snapshot is part of the operation rather than a step you meant to do first.
how it works →Backups that have been restored
Verification restores into a throwaway instance and checks it. “Fairly confident” becomes a status you can look at.
how it works →A record of what you did
Six months later, the reason that container has that flag is in the job that set it, not in your memory.
how it works →THE OUTCOME
What you get back.
The confidence to change things. Most homelabs ossify because the cost of touching anything is an unknown amount of a weekend, and the thing that makes it unknown is not having a record of how it got this way.
It runs comfortably on hardware you already have. Jobs execute on the target hosts, so this does not grow as you add machines.
OTHER WAYS PEOPLE RUN IT
Add the machine you understand least.
Discovery is read-only and takes about ten seconds. Start there.