10. Backups and restore
Open your project on the dashboard, then the deployment, and use the Backups section.
How backups work
| Kind | When it is taken | Kept until |
|---|---|---|
| Scheduled | At the interval you set | Removed when it is older than your Retain count |
| Manual | When you select Back up now | It does not count against your retain count |
| Before a restore | Automatically, before Restore in place | It does not count against your retain count |
Backups are taken from the second copy when it is up to date, so they do not slow down the copy serving your writes. Each backup is a full copy of the database. Restoring to a chosen moment in time is not offered.
A scheduled backup is skipped if the previous one is still running, or while another change such as a plan change is in progress. The next one runs as normal.
The Backups section
| Item | What it does |
|---|---|
| Back up now | Takes a manual backup. The database must be Running. Take one before every schema change |
| Enabled / Disabled | Shows whether scheduled backups run. Select it to switch them on or off, then confirm. Turning them off keeps the backups already taken |
| Edit settings | Change the interval and the retain count, then select Save |
| Stored | How much backup storage you are using |
| Free Allowance | Backup storage included in your plan, equal to the plan's storage size |
| Est. Monthly | The estimated monthly cost of storage beyond the allowance, in credits |
| Backup list | Started, State, Trigger, Size, Expires and Actions for each backup |
Download a backup
Select Actions, then Download. The download link is valid for 15 minutes. The file is an ordinary dump you can load anywhere: gzipped SQL for MySQL, or a pg_dump custom-format archive for PostgreSQL, which loads with pg_restore.
Restore into a new deployment
This is almost always the right choice. It gives you the old data without touching what is live.
- On the backup, select Actions, then Restore to new deployment.
- The create form opens, filled in from the source, with the name set to the source's name followed by
-restored. - Change what you need, then select Create database.
| Comes from the source and cannot be changed | Can be changed |
|---|---|
| Engine and version, topology, region, charset, collation and timezone | Name, plan, database name, username, password, backups, allowed list, cache and webhook |
If you leave Password blank, the new database uses the source's password. If you leave the allowed list empty, the source's list and its Allow linked API deployments setting are copied.
The new deployment is a full deployment. It uses one of the project's slots, it is billed from the moment it is created, and its credentials are shown once, as in Section 3.
Restore in place
This replaces all data in the live database with the backup.
- On the backup, select Actions, then Restore in place.
- Type the deployment's name exactly.
- Select Restore.
The database must be Running, and the backup must be from the same engine and version. A backup is taken automatically before the restore starts. If the restore was a mistake, restore that backup to get back to where you were.

