---
title: "Manual Backup"
url: "https://docs.chinaski.net/en/backup/manual-backup/"
type: "page"
language: "en"
description: "Take a point-in-time Chinaski backup on demand from the Backups page: what the .tar.gz contains, how it is verified, and how it is stored."
updated: "2026-09-29"
generated: "Markdown rendition of the HTML page at url, built by Chinaski"
---

# Manual Backup

Automatic backups run on a schedule, but you can also take a point-in-time backup any time you want: before big content changes, before applying an update, or as an off-site copy to download and store elsewhere. Everything lives under **Backups** in the admin area.

## Creating one

Click **Create backup** on the Backups page. The archive builds in the background and, when finished, **your browser downloads the `.tar.gz` file directly**, named `chinaski_backup_<date>_<time>.tar.gz`.

- Contents: the database, the media library, themes, and site templates: a complete point-in-time capture.
- Verification: unless you turned verification off, the archive is checked (extract-and-validate) **before** it’s published; a failed verify keeps a bad archive from being stored.
- Publish then prune: the archive lands in the instance’s local `backups/` directory via an atomic rename, the listing never shows a half-written file, and the local retention policy applies (keep the last N backups and/or backups for N days), with every pruned file logged to the audit trail including the retention values used.

Two long-running edge cases are handled for you: creation runs against a two-minute watchdog (a wedged tar aborts with a “took too long” flashback rather than hanging the page), and a scheduled backup firing in the same second as a manual one can’t interleave; the staging names are process-id separated.

## Getting archives off the server

Any archive in the local list (not just one you just made) can be downloaded from the Backups page; the download route streams the file as an attachment. File names are validated against the exact backup-name pattern, so the download control only ever serves backup archives.

Space (DigitalOcean Spaces) copies are listed separately when configured; local files and Spaces copies are tracked independently (uploading to Spaces is a separate action on the local backups list).

## Cleaning up

- **Delete** removes a local archive (single click on its entry, audited with the filename in the record).
- Local retention is the same policy that applies to scheduled runs: **keep the last N** backups and/or **backups for N days**, both configurable in the Backups tab (0 = unlimited for either). Pruning runs after every backup creation, manual ones included, and each pruned file is recorded in the audit trail with the retention reason.
- Staged-but-unfinished uploads from a crashed creation are swept automatically (anything in staging older than an hour), so failed attempts don’t pollute the listing.

## What manual backups are for

The schedule covers the everyday drift; manual ones are for moments you want a guaranteed-matching pair: right before a deploy of new template files, before a restore dry-run experiment, or as the macOS-style “golden” archive you download and keep elsewhere. Since the archive is the same format the restore path consumes, a manual download doubles as an off-server recovery artifact; the disaster-recovery runbook consumes exactly this file.
