Versioning keeps old states in the same bucket, in the same project, with the same key – that is not a backup. Clusterward copies every bucket each night into a backup bucket with Object Lock, ideally in a Scaleway project of its own and another region. Each night is a restore point, kept 30 days by default and tested every month.
An S3 backup is a nightly copy of every bucket Clusterward manages into a backup bucket with Object Lock. You set the backup bucket up once on the S3 page; after that backups are on for every bucket – including buckets added later. Only what changed is copied, and deleted objects stay restorable for the retention period.
Delete the bucket, lose the key or lose the region, and the versions go with it. A backup lives somewhere else, is kept for a fixed time and cannot be removed by the thing it protects against. That is exactly what an ISO 27001 audit asks under A.8.13.
| Aspect | Versioning | S3 backup in Clusterward |
|---|---|---|
| Location | the same bucket, the same project | a backup bucket of its own, ideally another project and region |
| Who can delete | anyone with the bucket’s key | nobody with the backup key – it has no delete right |
| Region fails | original and versions gone | the copy in the other region |
| Retention | until someone cleans up | a fixed period, 7 to 365 days |
| Evidence | none | restore tests and a CSV for the audit |
Register a Scaleway project of its own as an S3 provider. Its key needs IAM Manager rights once.
S3 → Backups → “Set up backups”: provider, region, bucket name, lock and days. No region is preselected – you decide.
A bucket with Object Lock and versioning, retention, clean-up rules, a backup key of its own and a test object – then “Ready”.
Every bucket gets its run; it ends with a restore point listing every object of that night.
Both modes lock every copy until its retention ends. They differ in whether an administrator of the backup project can remove a copy earlier. “Change retention…” changes the days and switches governance to compliance; compliance can only be lengthened, never shortened or turned back.
| Lock | Who can delete a locked copy | When to pick it |
|---|---|---|
| Governance | nobody with the nightly backup key; an administrator of the backup project in the Scaleway console | the default – lets you honour a deletion request in the backups too |
| Compliance | nobody – not Clusterward, not an administrator, not Scaleway support – until the retention ends | when a rule demands copies that cannot be removed; deletion requests then take effect only at expiry |
Copied is what is new or changed in size or checksum. Unchanged objects refer to earlier nights.
An object deleted from the bucket stays restorable for the retention period.
Clusterward creates the nightly key for this alone; it can write, but delete nothing.
Objects up to the 5 TB S3 limit: over 2 GiB in parts, each part retried on its own.
“Bucket backup failed” on an error, “Bucket backup missing” when a bucket had no good backup for 36 hours.
If buckets sit in the backup’s region or project, the card names them – one outage would hit both.
“Backup history…” on the bucket lists every night as a restore point. “Restore into a new bucket…” copies that night into a new, private bucket – the running bucket is never touched. Every object comes back at exactly that night’s state, with its content type and metadata; just one folder if you like.
Most of the time it is one deleted file: “Download a file…” searches a night’s files by name or path and downloads one through a link that is valid for five minutes.
A backup only counts once a restore has worked. Once a month Clusterward reads a sample back from every bucket’s latest restore point and compares size and, where present, checksum. A test only reads and writes nowhere; “Test restore” starts one right away.
| What | How |
|---|---|
| Sample | the newest and the largest object (up to 100 MiB) plus random ones – up to 50 objects or 500 MiB |
| Check | size must match, checksum where present |
| Failure | “Bucket restore test failed” to your channels |
| Evidence | “Download evidence…”: a CSV of every backup, restore and test, up to 400 days |
| Record | every download is in the audit log |
How buckets, providers and keys are created is covered under Object Storage. Clusterward backs up databases, disks and earlier versions in other ways – the overview is Backups & recovery.
The backup evidence is also one of the reports of the compliance page, which shows bucket backups and restore tests as ISO 27001 checks.
Buckets, IAM keys per application and bucket policies.
Go to Object StorageDatabases, disks, buckets and earlier versions at a glance.
Go to BackupsLive checks, reports and the compliance pack for your audit.
Go to the compliance pageYour question isn’t here? Write to us – we usually reply the same working day.
Ask a questionWe set up a backup target with Object Lock, restore a night into a new bucket and download the evidence for your audit.
