Cloud-to-cloud copy
A Cloud Storage job reads a container or prefix in one storage target and writes it into another. Where it runs is your choice per job: an agent of the customer's relays the data, or the Caryvane Cloud Agent moves it without touching their hardware.
Create the job #
- Jobs → New Job → Cloud Storage. Choose where it runs: Your Agent (a machine relays source → destination; use this for a LAN share at either end) or Caryvane Cloud Agent.
- Source target and one or more container or container/prefix specs; browse to pick them. Copy mode Full or Incremental.
- Destination target and directory. Source and destination must be different targets; both must belong to the same customer.
- Schedule: a cron, or Continuous for sources that report changes (below).
Incremental copies #
Incremental skips objects that are unchanged at the destination, judged by size and the content hash where both providers report one. For Egnyte sources Caryvane goes further: it records a change-feed cursor after each run and next time asks Egnyte only for what changed since, so a large tenant is not re-listed every run. A first run, or a run after the cursor is lost, does a full listing and seeds it.
Keys at the destination mirror the source: <directory>/<relative path within the prefix>, so a copy of archive/2024/ lands as <directory>/2024/….
Continuous copies #
Where the source can tell Caryvane that an object changed, the job can run continuously and copy each change within moments. Sources that can:
| Source | How | Set-up |
|---|---|---|
| Azure Blob | Event Grid | Set up Event Grid on the target; Caryvane registers the subscription itself. Details |
| Amazon S3, Wasabi | SNS | You create a topic, let the bucket publish to it, and subscribe the console's webhook. Details |
| Egnyte | Egnyte webhook | Registered by Caryvane when the target is tested. Details |
Continuous cloud copies run on the Cloud Agent. Each incoming event queues a copy of just that object; events for objects outside the job's prefixes are ignored, so a job can never be steered to copy something it was not configured for. The webhook URL and secret are under Configure webhook on the target.
Version history and conflicts #
A cloud copy honours the job's Keep previous versions setting at the destination like any other job. Conflict policy applies to scheduled copies; a continuous copy always writes the current object over the stored one, because the source has just said it changed.
Limits #
- A local folder or SMB share can be either end only when the job runs on an agent that can reach it; the wizard refuses the Cloud Agent for those.
- Very large objects stream between S3-style and Azure providers with no staging; other providers stage each object through the agent's or worker's scratch space, one at a time.
- A cloud job counts against the customer's Cloud Agent allowance; see Billing.