What a node-local backup can do

What a Foxcore backup is

Backup is a job on the instance. The node stops the guest for the job, writes a local tarball, and starts the guest again when desired power is running. The archive stays on that node. It is not copied to object storage and it is not an off-site backup.

Restore latest clears the instance data directory and unpacks the newest listed archive. It does not let you pick an older job id, and files added after that archive was written do not survive the restore.

Node-local archives are not exportable today

Backups has no download control. backup.List and backup.Delete talk to the node, and nothing copies the tarball off it. Files cannot bridge the gap: it reads one file at a time and caps a read at 256 KB, and a world archive is larger than that.

So an archive protects against a bad config, a failed mod update, or a world edit you regret. It does not protect against losing the node itself. If the node is gone, the tarball on it is gone too.

What that means for a real plan

Foxcore cannot currently hand you a durable copy of a world it holds. Keep any copy you already have outside Foxcore — the world you started from, a local test copy, or an archive from a previous host — and label it with the date and pack version.

Do not treat a node-local archive as an exit plan. If you need one, read what the panel can and cannot export before you cancel.

1. Before a config or pack change, stop the guest and create a backup.
2. Wait until On-disk archives lists the tarball.
3. Treat that tarball as node-local: it cannot be downloaded.
4. Keep the copy you already hold outside Foxcore until a transfer path exists.
5. Do not destroy an instance while the only copy of a world is on its node.

There is no schedule yet

The panel does not run backups on a timer. You create each one, and you delete old archives yourself. That is deliberate for this wave: a schedule that silently fails is worse than a manual action that reports its job state.

If the archive list is unavailable because the node session is down, the panel shows UnavailablePanel. Job history from the control plane can still show what ran. Do not destroy an instance while you still need its volume.