TL;DR
SAP Note 3255746 prohibits customer and third-party applications from calling the RFC modules of SAP’s Operational Data Provisioning (ODP) Data Replication API. Since SAP’s June 9, 2026 security patch (Note 3748819, CVE-2026-44754), that prohibition is technically enforced: ODP-RFC calls from non-SAP subscribers are rejected. A temporary opt-out (Note 3731818) expires in December 2026.
Any Databricks pipeline fed by an ODP-RFC extractor (the Azure Data Factory SAP CDC connector, the ODP components in most ELT tools, or custom PyRFC code) stops receiving SAP deltas once the patch is active without the opt-out. Compliant paths include ODP over OData, SAP Business Data Cloud (BDC) Connect via Delta Sharing, Datasphere replication flows, and non-ODP table or extractor interfaces.
If you’re reading this in November 2026, you have roughly seven weeks.
What SAP Note 3255746 actually says
The note is titled “Unpermitted usage of ODP Data Replication APIs.” It has hardened in stages:
| Date | Milestone | What changed |
| Oct 17, 2022 | Note 3255746, v1 | ODP-RFC use by customer or third-party apps declared unsupported; modules intended for SAP-internal use |
| Feb 2, 2024 | v4 | Status moves from “unsupported” to “not permitted” for on-premise and private cloud ABAP sources |
| Jul 7, 2024 | Update | SAP points customers to Datasphere for SAP-to-third-party replication |
| Apr 13, 2026 | Note 3439624 | Self-assessment tool released to audit ODP-RFC consumers |
| Apr 21, 2026 | v11 | ODP-RFC defined as exclusively SAP-to-SAP; problems from non-compliant use are the customer’s sole responsibility; BDC and ODP-OData named as alternatives |
| Jun 9, 2026 | Note 3748819 (CVE-2026-44754) | Security patch adds caller identification; unpermitted callers rejected |
| Jun 9, 2026 | Note 3731818 | Temporary opt-out restores unrestricted calls, at the customer’s risk |
| Dec 31, 2026 | Opt-out ends | SAP’s user-group briefing lists the suspension feature as available until this date |
For more than two years the rule was contractual. Most teams kept running ODP-based pipelines because nothing technically stopped them. The June patch changed that. Compliance is now a property of your SAP system’s patch level, not of how you read a support note.
How the block works
SAP shipped enforcement as a security fix on component BC-BW-ODP, rated CVSS 6.6. The framing matters. SAP describes the defect as missing caller identification in the ODP-RFC modules, which let non-SAP applications use the API outside its intended purpose. The fix checks every ODP-RFC call against permitted subscriber type patterns and rejects the rest.
Three consequences follow for engineering leadership:
- It arrives through your normal patch cadence. Basis teams apply security notes as routine hygiene, and the restriction also lands through later support packages. An upgrade weekend can switch off your SAP feed without anyone in data engineering being consulted.
- The failure is loud at source and silent downstream. Extraction tools surface SAP message RSAR 051, “Unpermitted caller.” Your Databricks jobs see nothing at all.
- The opt-out is a security decision. SAP states that the identified risk persists while the opt-out is active and that reverting is at the customer’s risk. Your CISO, not your data platform team, should own that sign-off.
What breaks: five failure modes in Databricks pipelines
Most SAP-to-Databricks architectures share one shape. An extraction tool reads ODP delta queues and lands Parquet or CSV in ADLS, S3 or GCS. Auto Loader or Lakeflow Declarative Pipelines (formerly Delta Live Tables) then builds Bronze, Silver and Gold. Note 3255746 hits the first hop, but the damage travels through every layer.
1.Ingestion stops, and your jobs still go green
The ADF SAP CDC connector reads SAP extractors, ABAP CDS views, BW objects and SLT-exposed tables through ODP-RFC, so it is affected. So are the ODP components in tools such as Matillion, Qlik and Theobald, and any custom Spark or Python job that uses PyRFC to call the ODP replication function modules (the RODPS_REPL_* family).
When the upstream call is rejected, no new files land. Auto Loader processes zero files, streaming tables report a successful update, and Gold refreshes on stale data. If you monitor job status rather than freshness (maximum ingestion timestamp per source against its SLA), the first alert will be a CFO asking why the revenue dashboard hasn’t moved since the weekend.
2.Delta state belongs to the old subscriber
ODP delta tracking lives in SAP’s operational delta queue (ODQ), keyed to a subscriber. A new extraction method means a new subscription, and a new subscription means a new delta initialization. For high-volume objects such as ACDOCA (the Universal Journal) or MATDOC, that is a reload of billions of rows, with real load on application servers and a reconciliation window finance has to approve.
The common trap: a team selects a replacement, then discovers the full reload won’t fit before the opt-out expires. Size the re-initialization before you choose the tool.
3.Your MERGE logic encodes ODP semantics
Silver-layer MERGE or APPLY CHANGES logic built on ODP feeds usually depends on ODP change metadata: record-mode indicators for inserts, updates and deletes, and counters that sequence changes within a load. Table-level CDC, OData deltas and BDC data products each express change differently. A shared data product is a current-state view, not a change log.
Swap the source without rewriting the apply logic and you get duplicate keys, lost deletes, or SCD Type 2 history that silently forks. Treat the switch as a data-contract change and regression-test it against a reconciled snapshot.
4.Extractor business logic disappears
Many pipelines rely on SAP Business Content extractors, such as the 2LIS_* logistics and 0FI_* finance families, because they carry decades of embedded logic: before and after images, currency handling, document-flow rules. Moving to raw table reads keeps data flowing but leaves you to rebuild that logic in Spark. That takes months and creates subtle variances against SAP-reported figures, which is exactly what auditors notice.
5.Governance and lineage reset
Unity Catalog lineage, quality expectations and access policies hang on table names, schemas and landing paths. A new source means new Bronze objects and new column names. With BDC, it also means a shared catalog object instead of a managed table. Permissions, data-quality rules and semantic-layer definitions all need remapping, and audit teams will want evidence that the old ODP path is gone, not merely idle.
Also Read: End of Support for SAP BW 7.5
What does not break
Precision matters here, because overreacting is expensive too. The note restricts the ODP Data Replication API over RFC, not RFC as a protocol. Table reads, BAPIs, standard function modules and the classic DeltaQ interface for extractors remain usable. ODP exposed through OData is the compliant variant SAP itself names, and data products shared through BDC are SAP’s sanctioned route out.
So the first job is an inventory, not a migration.
Your compliant options
| Option | How it reaches Databricks | Strengths | Trade-offs |
| ODP via OData | ELT tool or custom client pulls OData deltas into cloud storage | Keeps ODP delta semantics and existing extractors; SAP-sanctioned | Lower throughput than RFC; Basis must expose an OData service per source object; heavier application-server load |
| SAP BDC Connect for Databricks | Zero-copy Delta Sharing into Unity Catalog, with semantic metadata (names, descriptions, keys) synced automatically | No extraction pipeline to run; bidirectional; governed | Covers SAP-managed data products, not every custom table; BDC licensing; current-state shares, not change logs |
| SAP Datasphere replication flows | Datasphere, an officially supported ODP consumer, writes to ADLS, S3 or GCS | Retains ODP-based deltas; supported by SAP | Adds a licensed hop and another platform to operate |
| Non-ODP table or DeltaQ extraction | Tools using table CDC, BAPI or DeltaQ interfaces | Minimal SAP-side change; handles high volume | Incremental logic may move outside SAP; verify each vendor’s compliance claim against your Note 3439624 results |
Most enterprises will land on a portfolio: BDC data products for standard finance and supply-chain domains, OData or DeltaQ for custom Z-objects, and table-level CDC for a handful of very high-volume tables.
A seven-week plan of action
- Week 1, inventory. Have Basis implement Note 3439624 via SNOTE and run the self-assessment across ECC, S/4HANA and BW. Map every ODP-RFC subscriber to the Databricks tables it feeds.
- Week 1, remove patch surprises. Agree with Basis and security exactly when Note 3748819 and related support packages reach each system, and whether Note 3731818 is applied as a bridge.
- Week 2, monitor freshness. Alert on data age per SAP source in Databricks, not on job success. It is the cheapest insurance you will buy this quarter.
- Weeks 2 to 3, choose a target per object. Score each source on volume, change rate, depth of extractor logic and BDC data-product coverage.
- Weeks 3 to 6, migrate the top tier in parallel. Run old and new feeds side by side, reconcile row counts and financial totals in Silver, then cut over.
- Week 7, decommission and evidence. Remove ODP-RFC subscriptions in ODQMON, retire service credentials, and file the self-assessment output as audit evidence.
If the top tier cannot finish by December 31, decide now, with your CISO and SAP account team, whether to hold the patch on specific systems, and document that risk acceptance.
The strategic read
Note 3255746 is not an isolated support note. It sits alongside SAP’s clean-core and Business Data Cloud strategy, in which data leaves SAP through governed, SAP-controlled channels. SAP’s May 2026 acquisition of Dremio and its November 2026 announcement of native Apache Iceberg support and an Apache Polaris–based catalog in BDC point the same way: open formats, on SAP’s terms.
For data leaders the lesson is architectural. Treat SAP extraction as a contract you do not own. Isolate it behind a thin ingestion layer, and keep everything downstream (Spark compute, transformations, semantic models) portable across engines. Then the next policy change costs you a connector swap, not a re-platform.
Modak’s data engineering teams help enterprises assess ODP exposure, re-engineer SAP ingestion and reconcile cut-overs on Databricks and other Spark platforms. Talk to us before your Basis team’s next patch weekend.
Frequently asked questions
What is SAP Note 3255746?
It is SAP’s support note “Unpermitted usage of ODP Data Replication APIs.” It states that customer and third-party applications may not use the RFC modules of the ODP Data Replication API to extract data from SAP ABAP systems, on-premise or in private cloud. Version 11, released April 21, 2026, is the current version.
Does SAP Note 3255746 affect Databricks?
It does not affect Databricks itself. It affects any SAP-to-Databricks pipeline whose extraction layer uses ODP-RFC, including the Azure Data Factory SAP CDC connector and the ODP components in many ELT tools. Once the June 2026 security patch is active, those pipelines stop delivering SAP data.
What is the deadline for SAP Note 3255746?
Enforcement began with SAP’s June 9, 2026 security patch. A temporary opt-out under SAP Note 3731818 allows unrestricted ODP-RFC calls until December 31, 2026, at the customer’s own risk.
What error do blocked SAP pipelines show?
After the patch, extraction tools report SAP message RSAR 051, “Unpermitted caller.”
Are RFC_READ_TABLE and ODP via OData still allowed?
Yes. The note restricts ODP over RFC only. Table reads, BAPIs and DeltaQ remain usable, and SAP names ODP-OData as a compliant alternative.
What is SAP’s recommended path into Databricks?
SAP Business Data Cloud Connect for Databricks, which shares SAP data products into Unity Catalog through zero-copy Delta Sharing.


