A website, an API, a database or a bot is code, but above all it is whatever the application produces while running: a database, files uploaded by your users, accounts, content. If that disappears, a git push does not bring it back.
So here is what happens at Bot-Hosting: your data exists in two separate places.
- On your deployment’s server, where your application reads and writes its files.
- In an encrypted copy hosted by OVHcloud in Gravelines, France, in storage separate from the server.
If the first one is lost, the second lets us put everything back. This article explains how that copy is made, what it keeps, for how long, and what it does not guarantee.
How the copy is made
A small backup program runs on every server. It watches the files of each deployment, whatever it is (website, API, database, bot), and only sends a backup when something changed. A deployment that wrote nothing costs almost nothing, and when it does write, only the difference is sent.
Each deployment has its own backup repository, tied to the deployment and not to the server. When a deployment is moved from one server to another, it keeps its whole history.
Databases are backed up properly
Copying the files of a running database can produce an unusable copy. For database deployments, a consistent export is therefore taken inside the container before the files are backed up: mysqldump for MySQL and MariaDB, pg_dumpall for PostgreSQL, mongodump for MongoDB. Redis already writes its file atomically, so it needs no export. The export is limited to 10 GB uncompressed and 10 minutes; beyond that it is only retried once a day.
Hosted in France, encrypted before it leaves
Copies are stored in OVHcloud’s Object Storage service, in Gravelines. They are encrypted with AES-256 by restic, an open source backup tool, before leaving the server. The storage only ever receives encrypted data.
How long we keep it, and how often
Frequency and retention depend on your plan:
| Plan | Check frequency | History kept |
|---|---|---|
| Free | every hour | 1 day |
| Starter | 15 minutes | 3 days |
| Starter+ | 15 minutes | 7 days |
| Advanced | 5 minutes | 14 days |
| Pro | 1 minute | 30 days |
| Pro+ | 1 minute | 60 days |
| Enterprise | 1 minute | 90 days |
“Check frequency” is how often we look for changed files. A backup is only created when there is something new.
If you delete a deployment, it stays recoverable for 30 days, whatever your plan.
It is verified, not just stored
A backup you have never restored is a hypothesis. So:
- Every week, three random deployments are verified: an integrity check of the repository, then a real restore of the latest state into a scratch folder, compared with the file count and size on record. It is the same path a recovery after losing a server takes.
- Every day, we check that each deployment’s repository is still in the storage. If it is gone, it is uploaded again in full.
- If an upload is interrupted, even halfway through, the next one picks up and the repository stays consistent. This is tested.
- Failures raise alerts on our side, with growing retry delays.
What if a server goes down?
Your deployment can be restored onto another server from its latest backed-up state. The same mechanism powers planned moves between servers: the application is only stopped for a few seconds, the time it takes to copy the last changes.
What this copy does not promise
We would rather be clear than display a percentage:
- It is not a 100% guarantee. It is a second copy, separate from the first, which greatly reduces the risk of losing everything without removing it.
- Backups are in a single place. They are hosted in Gravelines, with no second region. Losing the storage would cost the history, not your deployments: the files on the server would be uploaded again within 24 hours.- History is limited by your plan. A file deleted longer ago than your retention period can no longer be recovered.
If your data is critical, keep your own copy too, for example a git repository for the code and an occasional export of your database. Our backups are there so an incident does not become a disaster, not to replace your own habits.
In short
Your websites, databases and bots run on your server, and an encrypted copy, verified every week by a real restore, is kept in France at OVHcloud. Two separate locations, constant checks and a retention period that fits your plan: your data gets a very high level of protection, so you can build and grow with confidence.
To see the frequency and retention of your plan, head to the pricing page.