Amazon S3
Amazon S3 is the reference object store — every S3-compatible provider below behaves like it. Caryvane streams to it, uses the bucket's own versioning when it is switched on, and can be told of changes through SNS.
What you need #
- An IAM user or role with
s3:ListBucket,s3:GetObject,s3:PutObject,s3:DeleteObjectands3:GetBucketVersioningon the bucket (adds3:PutBucketVersioningif you want Caryvane to switch versioning on for you). - The Access key ID and Secret access key.
- The bucket's region (e.g.
eu-west-2) and, optionally, a default bucket name.
Add it as a storage target #
- Storage → Add → Amazon S3. Name it for the customer, not the provider ("Henderson backups", not "S3").
- Enter the key, secret and region. The default bucket saves typing in the job wizard; jobs can still target any bucket the key can reach.
- Test. Caryvane lists the buckets it can see. If the list is empty the key has no
s3:ListAllMyBuckets; that is fine as long as the bucket name is typed into the job.
Test on the target's row proves the credentials; a target that fails shows the provider's own message until it is fixed. From then on it can be a job's destination, a cloud-to-cloud source, a cloud drive, or an archive.
What it supports #
| Capability | Amazon S3 |
|---|---|
| Streams uploads (no staging on the agent) | Yes |
| Version history | Native bucket versioning when the bucket has it on; otherwise a sidecar copy |
| Realtime cloud-copy source | Yes — via SNS, set up by you |
| Cloud-drive mount | Yes |
| Reachable from the Cloud Agent | Yes |
| Browse in the console | Yes |
Things to know #
Versioning. If the bucket has versioning enabled, Keep previous versions on a job costs nothing extra to configure — S3 keeps every overwritten object itself and Caryvane prunes past the limit. Caryvane will try to enable it when a job asks for versions and the key allows; otherwise old copies go to a .versions sidecar beside the data.
Realtime. To copy from a bucket the moment objects land, the bucket must publish to an SNS topic that Caryvane subscribes to: create a topic, allow the bucket to publish to it, point the bucket's notification configuration at it, then subscribe the console's webhook URL (shown under Configure webhook on the target). AWS signs every SNS message, so nothing else needs configuring. Full steps are in Cloud-to-cloud copy.
Multipart. Large objects go up in 16 MB parts (scaled up for very large files), so a dropped connection costs one part, not the file.
Keys. Files are stored under the job's directory then the source path namespaced by drive, e.g. backups/C/Users/jo/Documents/report.docx. Object keys are case-sensitive; Windows paths are not, and Caryvane keeps the case it read.