> ## Documentation Index
> Fetch the complete documentation index at: https://schedy.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Backup & Restore

> Take a safe online snapshot of Schedy's store and restore it, using BadgerDB's native backup - no dependencies, no downtime.

Schedy persists every task to the `data/` directory (BadgerDB).
That directory is a live LSM tree, so copying it while Schedy is running - `cp -r data/`, a volume snapshot mid-write, `rsync` of an open store - can capture a half-written state and restore to silent corruption.

Use the backup endpoint instead. It streams a consistent snapshot that is safe to take while Schedy is serving traffic.

## Take a backup

`GET /admin/backup` streams a full snapshot of the store. It sits behind the same API key as every other endpoint, so send `X-API-Key` if one is configured.

```bash theme={null}
curl -H "X-API-Key: your-secret" \
  http://localhost:8080/admin/backup \
  -o schedy-backup.badger
```

The response is a single binary file. Store it wherever you keep backups; run the command on a schedule (cron, a Kubernetes `CronJob`) for point-in-time snapshots.

<Note>
  The snapshot is a full copy each time - there is no incremental or `since` mode. For most Schedy stores (pending tasks plus recent history) that is a small file.
</Note>

## Restore a backup

Restore is an **offline** operation: it runs against a stopped server with the `restore` subcommand.

```bash theme={null}
schedy restore schedy-backup.badger
```

It loads the snapshot into the `data/` directory and exits.
To protect a running deployment, restore **refuses to run if `data/` already contains data** - it will never half-overwrite a live store:

```
restore: refusing to restore: data directory "data" is not empty
```

To restore, point Schedy at an empty directory (or move the existing `data/` aside first), run `schedy restore`, then start the server as usual.

<Warning>
  Nothing else may hold the data directory open during a restore. Stop the server first.
</Warning>