Skip to main content

Backing up to S3-compatible cloud storage

Amazon S3, Backblaze B2, Wasabi, Cloudflare R2, MinIO or any S3-compatible bucket: the profile, the folder in the bucket, the storage class, and the limits to know.

Updated
On this page

A Cloud destination is your own bucket at a storage provider you pay directly; Genie9 never holds your data. Anything that speaks the S3 protocol works: Amazon S3, Backblaze B2, Wasabi, Cloudflare R2, MinIO on a server of your own, and most others. Cloud destinations are part of Pro.

Before you start#

From your provider you need the bucket name, its region (or, for Cloudflare R2, your account ID), and an access key ID and secret access key that can list, read, write and delete in that bucket. Only a service GBM AI does not list by name needs its endpoint.

Step 1: Create an S3 profile#

A profile holds the keys, so several jobs can share one account.

  1. Open Settings ▸ Network & credentials, and under S3 profiles click Add profile. (Or click New profile… on the wizard's Where step.)
  2. Provider: Amazon S3, Backblaze B2, Wasabi, Cloudflare R2, or Other S3-compatible for MinIO and the rest.
  3. Profile name: anything, for example Wasabi EU.
  4. What the provider needs:
    • Amazon S3: the bucket's Region, for example eu-west-1.
    • Backblaze B2: the Region in your bucket's endpoint, for example us-west-004. GBM AI makes the endpoint from it.
    • Wasabi: the bucket's Region, for example eu-central-2. GBM AI makes the endpoint from it.
    • Cloudflare R2: your Account ID, 32 letters and digits, in Cloudflare's dashboard under R2. The region is automatic.
    • Other S3-compatible: the service's Endpoint, for example minio.example.com:9000, and its Region if it has one.
  5. Fill in Access key ID and Secret access key. The secret is sent to the agent once and stored encrypted on this PC; it is never shown again.
  6. Click Test. Signed in (210 ms) means the keys work.
  7. Click Save. The list shows each profile's provider and region.

S3 profile dialog - a Wasabi accountS3 profile dialog - a Wasabi account

Step 2: The Cloud destination in the wizard#

  1. On the Where step choose Cloud (S3-compatible).
  2. S3 profile: pick the profile you made.
  3. Bucket: the bucket's name.
  4. Folder in the bucket: it starts as the job's name. Leave it unless you share the bucket. The full s3:// address is shown under it.
  5. Storage class: see the table below. Amazon S3 offers all three; Cloudflare R2 offers Standard and Infrequent Access. Other providers have no choice here: the bucket's own default applies.
  6. Server-side encryption: The bucket’s default, S3-managed keys (SSE-S3) or AWS KMS key (SSE-KMS) on Amazon S3; The bucket’s default or Backblaze's own keys on Backblaze B2. Other providers encrypt at rest by themselves, so there is no choice here. This is the provider encrypting at rest, separate from the job's own AES-256 password; you can use both.
  7. Bandwidth limit: in Mbit/s, 0 for unlimited.
  8. Click Test destination. It checks the bucket, writes a small probe object and deletes it. Reachable and writable (180 ms) means the job is ready.
Storage classThe app's hint
Standard (the bucket’s default)Whatever the bucket is set to. Restores are immediate.
Infrequent AccessCheaper to keep, a small fee per read. Restores are immediate; best for sets you rarely open.
Glacier Instant RetrievalCheapest to keep, a higher fee per read, a 90-day minimum charge per object. Restores are still immediate.

What is different in the cloud#

  • Files are uploaded one by one as the backup runs. The set's manifest and catalogue are kept on this PC and copied into the bucket beside the files.
  • Small files are packed (on by default): files under 256 KB are stored together in pack files, so thousands of small files cost dozens of requests, not thousands.
  • Verification reads everything back. Verify files after each run downloads every stored object and checks it against the manifest, as for any destination. Providers that charge for reads or downloads charge for that too. For a large cloud job where that matters, turn Verify files after each run off on the Settings step.
  • Restores download from the bucket, at your provider's download price if it has one. Find keeps working from the index on this PC.
  • Watch the cost. A job with long retention on Standard storage grows steadily. Standard for the daily job and Infrequent Access for a weekly archive job is a reasonable split.
  • A bucket with versioning on keeps what retention deletes as old versions, still billed. Add a lifecycle rule at your provider that expires noncurrent versions, or use a bucket without versioning. A bucket with Object Lock is handled by GBM AI: see Locking backups in an S3 bucket.

Note

The bucket holds everything needed to restore without the PC that made the backups: on another PC, open them with Restore ▸ Open backups… ▸ An S3 bucket… and the same keys. See Restoring backups made on another PC.

Providers, briefly#

ProviderWhat you typeNote
Amazon S3the region
Backblaze B2the region from the bucket's endpoint (us-west-004)use an application key, not the master key (B2's S3 service refuses it); bucket names may have capital letters; Object Lock works
Wasabithe bucket's regionObject Lock works
Cloudflare R2your account IDno egress fees; R2 has no S3 Object Lock, so a job cannot lock its backups there
MinIO and other own serversOther S3-compatible: your server's host and portpath-style addressing is used automatically for any custom endpoint
Oracle Cloud Object StorageOther S3-compatible: <namespace>.compat.objectstorage.<region>.oraclecloud.comthe region is read from the endpoint; the access key is a Customer Secret Key from your Oracle user's settings
Alibaba Cloud OSS, Tencent COSOther S3-compatible: the bucket's region endpoint (oss-cn-hangzhou.aliyuncs.com, cos.ap-guangzhou.myqcloud.com)the bucket's name goes in the host name automatically, as these services require; a Tencent bucket's name ends in your APPID
Scaleway, DigitalOcean Spaces, Akamai (Linode), SakuraOther S3-compatible: the region's endpoint (s3.fr-par.scw.cloud, nyc3.digitaloceanspaces.com, us-east-1.linodeobjects.com, s3.isk01.sakurastorage.jp)the region is read from the endpoint where the service needs it
Google Cloud Storage with HMAC keysOther S3-compatible: storage.googleapis.comworks, but the File server or cloud destination with a service account key is the better way to Google Cloud

SFTP, FTP, WebDAV (Nextcloud, ownCloud), Azure Blob Storage and Google Cloud Storage are the File server or cloud destination since 10.6: see Backing up to a file server or business cloud. Backblaze's own (non-S3) API is not supported; use B2's S3 endpoint, as above.

Troubleshooting#

"The service refused the keys"#

The test says why, in its own words:

  • The service does not know this access key: the key ID is mistyped, or the key was deleted. On Backblaze B2 it is also what the master application key gets: make an application key and use that.
  • The secret key does not belong to this access key: the secret is wrong. Copy it again; most providers show it only once, so make a new key if you no longer have it.
  • The service refuses the account: the account is closed or suspended, or its free trial ended. Sign in at the provider and check the account.
  • S3 refused the credentials (AccessDenied): the key works, but has no rights on this bucket. Make a key with list, read, write and delete on that bucket.

E-DEST-006 when a job runs#

The S3 profile this job uses isn't on this PC. The job was imported, or its profile removed. Open the job and pick or create the profile that holds the keys for its bucket.

Was this article helpful?