Use Data Control Tower (DCT) to upgrade Continuous Data and Continuous Compliance engines without signing in to each engine's SysAdmin UI. DCT stores an upgrade image once, transfers it to the engines you choose, runs each engine's upgrade verification checks, and applies the upgrade when you initiate it.
How an upgrade works
Each run is an upgrade operation. DCT reports progress for every engine through these phases:
Transfer: Creating an operation starts the transfer. DCT copies the image to each engine and waits for it to unpack, which typically takes 15 to 30 minutes. This phase does not cause VDB downtime.
Verify: After transfer, the engine automatically runs its upgrade checks (SUV), and DCT captures the results. This phase does not cause VDB downtime.
Apply: You initiate this phase after reviewing the checks. The engine installs the new version and, for Upgrade and Reboot, reboots. VDBs are unavailable during the reboot.
Transfer and Verify can run before your maintenance window. Run Apply during the maintenance window. Through the API, you can apply to a subset of engines at a time for a rolling upgrade. DCT also runs a post-upgrade validation to confirm that the engine is running the new version.
Before you begin
DCT must be running version 2026.5, with Delphix telemetry set up and active. For general preview activation requirements, see Enable feature flags.
The Continuous Data or Continuous Compliance engines must be registered in DCT and running version 2026.4.0.0 or later. Upgrade engines on an earlier version directly to 2026.4.0.0 or later first.
The upgrade image must be an engine upgrade file with a
.taror.tar.gzextension. Its version must be newer than the engine's version, and the engine must meet the image'smin_source_versionrequirement.Upgrade replication targets before sources. A replication source cannot move ahead of its targets.
Your role must be Admin or DevOps, or a custom role with Upgrade Image permissions for Create, Read, and Delete; Upgrade Operation permissions for Create, Read, Update, and Validate; and Read permission on the target engines.
The DCT upload volume must have free space equal to at least twice the image size while the upload is assembled. Otherwise, the upload fails with Insufficient disk space.
The ingress or load balancer in front of DCT must accept request bodies of at least 550 MB. The UI uploads in 50 MB chunks, and the API accepts requests up to 550 MB. A smaller request-body limit can reject chunks with an HTTP 413 error.
DCT does not provide an ingress resource for uploads. Configure the request-body limit on your ingress controller or load balancer. For NGINX Ingress, add this annotation to your Ingress resource:
nginx.ingress.kubernetes.io/proxy-body-size: 550m
For other ingress controllers and load balancers, refer to your platform's documentation. For DCT ingress configuration, see Ingress setup.
Enable the preview
Until the preview is enabled, upgrade API calls return 403 Engine upgrade feature is not enabled, and the upgrade pages do not appear in DCT. An administrator enables the preview once for the DCT deployment.
The examples use a shell with curl. Replace <dct-host> with your DCT hostname and <api-key> with an API key for a user with the required permissions:
DCT=https://<dct-host>/dct/v3
AUTH="Authorization: apk <api-key>"
Confirm that Delphix telemetry is active.
Read the current feature flags and record the
feature_flagslist:curl -s "$DCT/management/properties" -H "$AUTH"Add
ENGINE_UPGRADEto the list and send the full list back. The update replaces the entire list, so include every existing flag. This example is for a deployment with no existing flags:curl -s -X PATCH "$DCT/management/properties" -H "$AUTH" \ -H "Content-Type: application/json" \ -d '{"feature_flags": ["ENGINE_UPGRADE"]}'Wait about 15 seconds. No restart is required.
Sign in to DCT or refresh the page. Confirm that Admin > Engine Management > Upgrade Images is available. If API calls still return the 403 message, read the flags again and confirm that
ENGINE_UPGRADEis present.
To disable the preview, update the full feature flag list without ENGINE_UPGRADE.
Upgrade engines in the UI
Upload an upgrade image
Navigate to Admin > Engine Management > Upgrade Images.
Click Upload and choose the upgrade image file. The upload starts immediately. Keep the dialog open until it finishes.
Wait for the image status to become Ready. DCT assembles the image, verifies its checksum, and reads its version information after the last chunk arrives.
The table shows each image's Version, Size, Source, Downloaded date, and Status. If an upload is interrupted, uploading the same file again resumes from the last chunk DCT received.
Uploading: DCT is receiving the image. Click Cancel Upload to stop.
Processing: DCT is assembling and verifying the image.
Ready: The image can be used, and Upgrade Engines is available.
Invalid: Verification failed. The file might be corrupt, not an engine upgrade image, or a duplicate of an uploaded image. Delete it and upload a valid image.
File Missing or Failed: The image cannot be used. Delete it and upload it again.
Upload errors include Insufficient disk space, Image already exists, and Chunk too large. For Chunk too large, check the request-body limit described above.
Start an upgrade
On Upgrade Images, click Upgrade Engines for a Ready image. The wizard opens on Select Engines, with the target version taken from the image.
Under Choose upgrade type, choose a type. It applies to every engine in the operation and cannot be changed later:
Upgrade and Reboot is the default and recommended type for the preview. It corresponds to
FULLin the API. The engine upgrades and reboots, making VDBs unavailable during the reboot.Upgrade only corresponds to
DEFERREDin the API. It upgrades only the management stack, without a reboot or VDB downtime. The engine continues running the previous version until rebooted. DCT cannot finish a deferred upgrade in this release; finish it on the engine itself.
Choose the engines to upgrade. Only eligible engines are listed, with Name, Type, Current Version, and Status. If none is eligible, the wizard shows No eligible engines found for this image.
Click Next. DCT validates the replication topology. If the selection is invalid, the wizard lists the affected engines and reasons. Adjust the selection and click Next again.
On Confirm & Initiate, review the Upgrade Summary and each engine's current and target versions.
Click Start Upgrade. Transfer and Verify run in the background. An engine can belong to only one active upgrade operation at a time. If the operation cannot start, the wizard stays open and displays an error.
Monitor progress and resolve checks
Navigate to Admin > Operations and click the Upgrade Engine Operation job. The upgrade details page refreshes automatically and shows overall status, start time, run time, and Total Progress.
Expand a phase to view each engine's progress and state. Phases include Transferring upgrade image, Running pre-upgrade verification, Resolving upgrade checks, Applying upgrade, and Post-upgrade validation.
When Verify finishes, the operation changes to Action Required. Expand each engine in Upgrade Checks Require Attention to review its checks. Severities are CRITICAL, WARNING, and INFO. Informational checks need no action.
Correct the underlying issue on the engine, then click Resolve for the check. Alternatively, click Ignore only if you accept the risk of leaving the issue uncorrected.
Repeat until no CRITICAL check remains unresolved. Apply Upgrade is unavailable while any engine has an unresolved CRITICAL check and identifies the blocking engines. When nothing remains to resolve, the page shows All pre-upgrade checks passed. Ready to apply the upgrade.
The overall operation status can be Running, Action Required, Completed, Failed, or Canceled.
Apply, cancel, or terminate an upgrade
During your maintenance window, click Apply Upgrade when no CRITICAL checks remain unresolved.
Monitor Applying upgrade. Each engine moves from Upgrading to Upgraded. With Upgrade and Reboot, VDBs are unavailable until the engine comes back online. To apply to only a subset of engines, use the API.
While the operation is Action Required, click Cancel Upgrade to end it before any engine is upgraded. You cannot remove an individual engine from an operation. To exclude an engine, cancel while the operation is Action Required and start a new operation without that engine.
While the operation is Running, click Terminate to stop it. Terminate is also available after an apply failure if the operation has not ended. Recovery actions are not available in the UI during this preview. For API recovery options, see the recovery section below.
Delete an upgrade image
Images remain in DCT until you delete them. To free disk space, click Delete Image on Upgrade Images and confirm. Deletion cannot be undone. During an upload, the action is labeled Cancel Upload.
Upgrade engines through the API
The API supports the UI workflow plus subset and rolling Apply and engine recovery. Endpoints are grouped under /upgrade in the Swagger UI. Enable the preview first, and use the DCT and AUTH variables defined above. Replace all angle-bracketed placeholders with your values.
1. Upload an image
Upload images in ordered chunks, using the original filename for each chunk. Requests are capped at 550 MB; 50 MB chunks are recommended. This example requires a Bash-compatible shell with GNU stat and split. Run it in a directory without other files named part_*. Check each response before proceeding to the next chunk; stop and resolve any upload error.
FILE=<engine-upgrade-image>.tar.gz
CHUNK=$((50*1024*1024)); TOTAL=$(stat -c%s "$FILE")
split -b $CHUNK -d -a 4 "$FILE" part_
n=0
for p in part_*; do
curl --fail-with-body -sS -X POST "$DCT/upgrade/images/upload" -H "$AUTH" \
-F "file=@$p;filename=$FILE" -F "_chunkNumber=$n" -F "_chunkSize=$CHUNK" \
-F "_currentChunkSize=$(stat -c%s "$p")" -F "_totalSize=$TOTAL" || break
n=$((n+1))
done
Intermediate chunks return HTTP 200. The final chunk returns HTTP 201 with upgrade_image.image_id and status PROCESSING.
After an interruption, GET /upgrade/images/upload?file=<filename> returns bytes_received. Resume at chunk number bytes_received / CHUNK. Upload sessions idle for more than one hour can be purged.
Poll GET /upgrade/images/<image_id> until status is AVAILABLE, meaning checksum and version verification succeeded. INVALID indicates a corrupt file, a non-upgrade image, or a duplicate image. The API status AVAILABLE corresponds to Ready in the UI.
2. Choose engines and validate replication topology
curl -s "$DCT/upgrade/images/<image_id>/eligible-engines" -H "$AUTH"
curl -s -X POST "$DCT/upgrade/images/<image_id>/prevalidation" -H "$AUTH" \
-H "Content-Type: application/json" \
-d '{"engine_ids": ["<engine-id>"]}'
The eligible-engines endpoint filters by version. Prevalidation checks replication topology and returns validated_engines and unsupported_engines, with a reason for each unsupported engine.
validated_engines.3. Create the operation
curl -s -X POST "$DCT/upgrade/operations" -H "$AUTH" \
-H "Content-Type: application/json" \
-d '{"image_id": "<image_id>", "engine_ids": ["<engine-id>"], "upgrade_type": "FULL"}'
Creating the operation starts Transfer and Verify. The response is a job whose id is the operation ID used as <op_id> in subsequent calls. Choose the upgrade type at creation; Apply uses this value. An engine can belong to only one active operation at a time.
4. Monitor progress
curl -s "$DCT/upgrade/operations/<op_id>/report" -H "$AUTH"
curl -s "$DCT/jobs/<op_id>" -H "$AUTH"
The report returns overall_progress from 0 to 100 and an entry for each engine and phase. Each entry includes stage (TRANSFER, VERIFY, APPLY, or ROLLBACK), status, message, and progress_percentage. The jobs endpoint returns the overall operation status.
Transfer automatically retries transient network failures with backoff for up to 14 days and allows 60 minutes for an engine to unpack the image. Wait until every engine's VERIFY entry is COMPLETED before reviewing checks.
5. Review and clear verification checks
curl -s "$DCT/upgrade/operations/<op_id>/verify-result" -H "$AUTH"
Each engine has a checks list containing severity (CRITICAL, WARNING, or INFORMATIONAL), status, impact, and action. checks: null means results have not been captured; an empty list means no issues were found.
Any CRITICAL check not marked RESOLVED or IGNORED blocks Apply. Fix the issue and mark the check resolved, or ignore it only if you accept the risk. Each request accepts up to 20 references, using each check's suv_reference value:
curl -s -X POST "$DCT/upgrade/operations/<op_id>/verify-result/resolve" -H "$AUTH" \
-H "Content-Type: application/json" \
-d '{"engine_id": "<engine-id>", "suv_references": ["<suv_reference>"]}'
To ignore checks, send the same body to POST /upgrade/operations/<op_id>/verify-result/ignore. The response reports SUCCEEDED or FAILED for each reference.
6. Apply during the maintenance window
curl -s -X POST "$DCT/upgrade/operations/<op_id>/apply" -H "$AUTH" \
-H "Content-Type: application/json" \
-d '{"engine_ids": ["<engine-id>"]}'
Specify all engines in the operation or a subset for a rolling upgrade. Apply is all-or-nothing for the engines in the request: if any has a blocking CRITICAL check, the call returns HTTP 409 with blocking checks per engine, and no engine is changed. HTTP 400 means the image is not yet verified on an engine.
Monitor progress through the report endpoint. The operation becomes COMPLETED when every engine has applied successfully.
Recover from an apply failure
If an engine fails while taking its data sources offline, it enters DISABLE_FAILED. These POST endpoints accept no request body and are available only for an engine in that state:
/upgrade/operations/<op_id>/engines/<engine_id>/retry: After correcting the cause, run Apply again normally./upgrade/operations/<op_id>/engines/<engine_id>/continue: Proceed with Apply while ignoring failures to quiesce sources. Use this only if you accept those failures./upgrade/operations/<op_id>/engines/<engine_id>/rollback: Return the engine to its previous version. A rollback ends the operation asFAILED.
HTTP 422 means the engine is not in DISABLE_FAILED. HTTP 409 means recovery is already running. If Apply fails for another reason, recovery calls return HTTP 422 and the operation stays open. Contact your Delphix Account Team.
Additional API endpoints
GET /upgrade/images: List upgrade images.GET /upgrade/images/<image_id>: Get an image.DELETE /upgrade/images/<image_id>: Delete an image.GET /upgrade/operations/<op_id>/verify-result/<engine_id>: Get verification results for one engine.
Limitations and known issues in DCT 2026.5
Only engines on version 2026.4.0.0 or later are eligible. There is no override. Engines on the same version as the image or a later version are not eligible.
All engines in an operation use the same image and upgrade type. An
upgrade_typevalue on the Apply call is ignored. DCT cannot initiateFINISH_DEFERREDin this release. UseFULLfor the preview, or finish aDEFERREDupgrade on the engine itself.API prevalidation is advisory and not enforced at operation creation. The UI runs it when you click Next.
GET /upgrade/operations/<op_id>/apply-resultsand its/<engine_id>variant appear in the API specification but return HTTP 501 Not Implemented. Use the report endpoint instead.Retry, continue, and rollback are API-only and require
DISABLE_FAILED. The UI does not provide recovery actions.You cannot remove a single engine from a running operation. Cancel while the operation is Action Required and start a new operation with the revised selection.
After an upload, Upgrade Engines might remain unavailable. Refresh the page to enable it.
View Details in the Operations list might open generic job details instead of upgrade details. The generic page does not show upgrade phases or checks. If you cannot reach upgrade details, contact your Delphix Account Team.
Feedback
Send feedback to your Delphix Account Team. Include the number of engines, upload and phase durations, clarity of verification checks and errors, fit with your maintenance workflow, automation experience, and any recovery limitations you encountered.