Version requirement: OCI offline replication, including WORM support, requires Continuous Data engines running 2026.6.0.0 or later and DCT 2026.6 or later. This functionality is not available in DCT 2026.5.
OCI offline replication transfers data between Continuous Data engines through an Oracle Cloud Infrastructure (OCI) Object Storage bucket. The source writes serialization points to the bucket, and the target reads them without a direct connection to the source. A bucket can have a write once, read many (WORM) retention rule to prevent changes to or deletion of stored objects during the retention period.
Use DCT replication policies to configure the send and receive profiles. This topic covers OCI storage; for NFS-based transfers, see Offline replication (air gap).
Before you begin
-
Use separate source and target Continuous Data engines running 2026.6.0.0 or later. The corresponding send and receive policies cannot use the same engine.
-
Use DCT 2026.6 or later, and confirm that the Create Replication Policy wizard offers OCI Object Store as a storage type. Your DCT user and Access Groups need permission to manage the policies and related datasets.
-
Configure a bucket that both engines can access. For immutable replication data, configure the WORM retention rule in OCI before sending data. The engine detects the bucket's retention rules; there is no WORM switch or retention-duration setting in the replication policy.
-
Grant the identities used by the policies permission to read, write, and list objects, manage multipart uploads, and read retention rules on the bucket. For instance-principal access, use an OCI dynamic group and IAM policy. For API key access, grant access through the OCI user's group and IAM policy.
-
Synchronize both engines' clocks using NTP. OCI can reject signed requests if the clock has drifted.
Choose an authentication method
The send and receive policies can use different authentication methods or identities against the same bucket.
-
Managed credentials (instance principal): Uses the engine VM's OCI identity. The engine must run in OCI and belong to a dynamic group with access to the bucket. Enter only the bucket name in the policy.
-
API key-pair: Uses an OCI user's API signing key. This method also supports engines outside OCI. Upload the public signing key to the OCI user, then enter the region, tenancy OCID, user OCID, fingerprint, and matching PEM-encoded RSA private key. Enter the optional passphrase if the private key is encrypted.
For key setup instructions, see Oracle API signing-key documentation. DCT forwards the private key and passphrase to the engine; it does not store or return them. The engine stores them encrypted and does not display them.
Create the send policy
-
In DCT, go to Data > Replication > Policies and click + Replication Policy.
-
On Traffic Mode, click Offline - Send, then click Next.
-
Complete Replication Type and Details for the replication you want to configure.
-
On Replication Target, choose OCI Object Store for Storage Type. Enter the OCI Bucket Name.
-
Choose the authentication method and enter its required fields as described above.
-
On Replication Objects, choose the objects to replicate from your intended source engine.
-
On Parameters, configure automated replication if you want scheduled sends. Optionally set Full Send Interval (runs), Full Send Interval Duration, or both. The runs interval must be an integer of at least one. The duration field takes a number and a unit, such as seven days.
For a bucket with finite retention rules, the duration must be shorter than the longest finite retention period. Saving the policy validates this interval against the bucket. A runs-only interval is not checked against retention. If you change the bucket's retention later, recheck the full-send interval yourself.
-
Review Summary and click Save Policy. Open the saved policy and record its Source Replication Tag from the Overview tab. The receive policy must use this tag.
A successful save does not prove that the credentials have all required bucket permissions. Required API key fields and the RSA PEM key are validated when the profile is saved. Some settings, including retention-interval validation, also require bucket access at save time. Other OCI access errors appear when the policy runs.
Create the receive policy
-
Start another replication policy. On Traffic Mode, click Offline - Receive, then click Next.
-
Complete Details.
-
On Replication Target, choose the Target Engine that will receive the data. It must be different from the source engine.
-
On Replication Target, choose OCI Object Store for Storage Type. Enter the same OCI Bucket Name used by the send policy.
-
Choose the authentication method for the target engine and enter its required fields. You can use the source's API key or a different key with access to the bucket.
-
Enter the send policy's tag in Source replication tag. For API key authentication, this field is below the key fields. The tag cannot be changed after profile creation. On a target engine, use only one receive profile for a given source tag.
-
On Parameters, configure automated replication if you want scheduled receives.
-
Review Summary and click Save Policy.
Send and receive data
-
Open the send policy and use its Actions menu to open the send dialog. Choose how this run should send:
-
Use configured policy: The engine decides whether to send a full or incremental serialization point from the full-send policy.
-
Full Send: Forces a full serialization point for this run.
-
Incremental Send: Forces an incremental send and discards an in-progress full send.
-
-
Click Send. Check Execution History and wait for the send to complete successfully before starting the receive.
-
Open the receive policy's receive dialog from Actions. Choose how this run should receive:
-
Use configured policy: Uses the receive profile's stored receive policy. The default is
NEXT_IN_CHAIN, which receives the next serialization point in sequence. -
Next in Chain: Receives the next serialization point in sequence.
-
Full Receive: Selects the latest available full serialization point ahead of the target, skipping intervening points. If none qualifies, it receives the next point in sequence. Subsequent receives apply the incremental points after the selected full point.
-
-
Click Receive. Check Execution History and confirm that the receive completes successfully. The first successful receive creates the replica dataset on the target.
The first send is full. Later sends are incremental unless the configured policy or a one-time override requires a full send. The send policy's Overview tab shows Last Full Send. The number of runs since the last full send is available through the DCT API as runs_since_last_full_send, but is not shown in the GUI.
Archive restore behavior
If objects are in OCI's Archive tier, the engine restores them automatically during the receive. The job reports WAITING_FOR_UNARCHIVE while waiting. Metadata and data can require separate restores, so a receive may wait more than once. Restore time depends on OCI. Do not submit a separate unarchive request.
Recover a target that has fallen behind
If a required serialization point has been deleted, the receive fails and identifies the missing point. Run a Full Send on the source, wait for it to complete, then run a Full Receive on the target. Continue receiving the subsequent incremental points normally.
Retention and cleanup
Keep a recent full serialization point and all subsequent incremental points needed by your targets. Configure full-send frequency and OCI cleanup so required data remains available. Expiry of the OCI WORM retention policy allows objects to be deleted; it does not by itself delete them or reclaim space.
The engine's receive-profile property cleanupOfflineData defaults to false and must remain disabled for a WORM bucket. Enabling it is rejected when a retention rule exists. If retention is added after cleanup was enabled, the cleanup step fails at run time. On the target engine, use the CLI or Engine API to set cleanupOfflineData=false on the receive profile before running it against the protected bucket. This is an engine-level setting, not a field in the DCT wizard described above. For engine-level profile configuration and cleanup guidance, see OCI WORM offline replication in Continuous Data.
Local send and receive caches are cleaned up separately after jobs. Do not delete bucket objects still needed by any target.
DCT API field reference
These fields map the wizard settings to the DCT API. They are not Engine API property names. For endpoint definitions and payloads, use the DCT API reference for your installed version.
The DCT API uses the ReplicationProfile resource. To trigger an on-demand full send, use POST /replication-profiles/{id}/force-full-send with the send profile's ID. This is a DCT API operation, not an Engine API endpoint.
| Setting | DCT API property |
|---|---|
| OCI storage type | offline_storage_type=OCI_OBJECT_STORE |
| Bucket name | oci_bucket_name |
| Authentication mode | oci_auth_mode: INSTANCE_PRINCIPAL (default) or API_KEY |
| Region and identity | oci_region, oci_tenancy_ocid, oci_user_ocid, oci_fingerprint |
| Signing key | oci_private_key, oci_passphrase |
| Full-send intervals | full_send_interval_runs, full_send_interval_duration |
| One-time execution overrides | send_mode, receive_policy |
API values for full_send_interval_duration use ISO 8601 duration format, such as P7D for seven days. The send overrides map to FULL and FORCE_INCREMENTAL. Receive overrides map to NEXT_IN_CHAIN and LATEST_FULL_PREFERRED. Omit the override to use the configured policy.
Rotate an API signing key
Upload the replacement public signing key to the OCI user before updating the profiles that use it. Through the DCT API, update oci_fingerprint, oci_private_key, and oci_passphrase if required on each affected profile. If the send and receive profiles share the key, update both before running them with the replacement key.
Updates can omit unchanged credential fields to keep the engine's stored values. Do not resubmit redacted credentials from a read response.