47 Known Issues
This page lists current limitations, bugs, and behavior inconsistencies identified in Oracle AI Data Platform.
These issues are actively being tracked and will be addressed in future updates.
When reporting any issues regarding AI Data Platform resources (e.g. Workspace, Compute), you should also share the ID of the resource. You can find the resource ID by going to the page where the resource is listed, clicking 
Actions and clicking Copy ID.
Table 47-1 Known Issues
| Resource | Category | Error and How to Find the Problem | Workaround | |
|---|---|---|---|---|
| 1 | AI Data Platform Instance | UPDATE | Updating tags on the AI Data Platform instance does not retroactively apply to sub resources that were already created in the same compartment by AI Data Platform. | Tag updates only apply to newly created resources in the customer tenancy after the tag change.
If consistent tag values are required across all resources, manually update tags on existing sub resources. |
| 2 | AI Data Platform Instance | UPDATE | When a user selects or creates an external table using a bucket that has already been tagged/associated with another AI Data Platform instance the service blocks reuse. Deleting the external table from the original instance does not allow reusing the same bucket in the new AI Data Platform instance. | No current workaround. Once a bucket is chosen and the external table is created, the association is fixed. Customers must provision a new bucket if they need to create a new external table with a different AI Data Platform instance and copy the data to this new location. |
| 3 | AI Data Platform Instance | CREATE | A single bucket location cannot be referenced by multiple AI Data Platform instances for external tables at the same time. Attempting to reuse the same bucket results in validation failure. | No current workaround. Each AI Data Platform instance requires its own dedicated bucket for external tables. Customers must create separate buckets per instance for creating new external tables in a new AI Data Platform instance. |
| 4 | AI Data Platform Instance | UPDATE | Cannot edit tag namespaces after instance creation. | No current workaround |
| 5 | Auto-populate | CREATE | Auto-populate creation fails if the bucket has already been tagged with another AI Data Platform instance. |
|
| 6 | Compute | UPDATE | Updating a cluster is currently only possible while it is active. Updating a stopped cluster is currently not possible. | To update a stopped cluster, you can either start the cluster and update it or delete and create that cluster again with the changes. |
| 7 | Workspace | MOVE | From a workspace page, moving multiple items by selecting multiple items at one go is currently not possible. | Users can move one item at a time by selecting that item first, then selecting Move from the Actions Menu. |
| 8 | External Catalog | REFRESH | When a user creates an external catalog, it runs a background job to harvest the metadata from the external source. The background job takes a long time for the metadata from the external system to reflect in the external catalog. Clicking on the refresh icon in the master catalog while the first job is still in progress causes problems in the processing of the background job. | You should not click the Refresh button until the job has finished. Users can see the progress of the background job in the History tab. When the job is complete, you can click Refresh if all the expected objects have not appeared in the external catalog. |
| 9 | External Catalog | REFRESH | When a user creates an external catalog, it runs a background job to harvest the metadata from the external source. If a column in a table in that external source has "$" in the name it will not be harvested. | No current workaround |
| 10 | External Catalog | CREATE | Users cannot create external catalogs with regional wallet files. | You can create external catalogs using instance wallets or provide the instance details and create an external catalog. |
| 11 | External Catalog | LIST | Users cannot list or use synonyms in an external catalog when using an AI Lakehouse source. | No current workaround |
| 12 | Data Ingestion | CREATE | Users cannot write to an Oracle Autonomous AI Lakehouse table using a notebook without first creating an external catalog | Users are expected in general to first create an external Autonomous AI Lakehouse catalog and then write it to an Autonomous AI Lakehouse table. |
| 13 | External/Managed Table | CREATE | Users cannot create external tables when the input data is in multi-line JSON format. | No current workaround |
| 14 | Data Sharing | CONSUMPTION | Users cannot receive a share using Delta Sharing protocol. | Users can load the shared data into a data frame using the delta-sharing library's load_as_spark(<<table_path>>). |
| 15 | Data Sharing | CONSUMPTION | Users cannot consume data shared by AI Data Platform in Autonomous AI Lakehouse | No current workaround |
| 16 | Auto Populate | CREATE | The Auto Populate feature does not support table creation when the target location contains delimited data files with a delimiter other than a comma. | No current workaround |
| 17 | Workflow | CREATE/SCHEDULE | Directly accessing OCI Object Storage (or GenAI service) is currently not supported for scheduled jobs | Configure external volumes for Object Storage access in the workflow job. There is no current workaround to access generative AI models |
| 18 | Workspace configured for Private Network | Access AWS S3 Buckets | For workspaces configured for Private Network, currently there is a known issue about AWS S3 access from Notebooks and Workflows. | File a support ticket if you need to access AWS S3 buckets from notebooks and workflows for Workspaces configured for private networks. Please indicate the workspace key in that ticket.
AI Data Platform team will enable a temporary workaround for your tenancy while we work on a permanent solution. |
| 19 | Compute | Monitor compute notebooks | All notebooks attached to a compute cluster as displayed are Active regardless of actual status | No current workaround
AI Data Platform team is working on a fix. Notebooks will display the correct status when the fix is deployed. |
| 20 | AI Data Platform Instance | Network source and Admin recovery | Network source based access control does not currently cover the admin recovery related actions. | No current workaround
AI Data Platform team is working on a fix. Until the fix is released, you can temporarily disable network source based access control, if you need to perform the infrequent actions related to admin recovery. |
| 21 | AI Data Platform Instance | Delta tables | Delta table writes and reads hit OCI Object Storage throttling errors if there are too many small files in the Delta tables. | No current workaround
AI Data Platform team is working on automatic capabilities related to compaction. Until the fix is released, we recommend using the Optimize command frequently to avoid throttling-related errors. |
| 22 | Compute | Edit cluster | Changing shapes (changing compute architectures) while editing Spark clusters is currently not supported. | You can create a new cluster if you need a different compute shape. We are evaluating use-cases where changing shapes are critical. Please file a support request describing your use-case. |
| 23 | Data lineage | Lineage capture | Data lineage is not currently generated for Apache Spark Structured Streaming tasks run from notebooks or Workflow jobs. As a result, the lineage graph might not display relationships between source and target datasets. | No current workaround |
| 24 | AI Data Platform | Browser support | AI Data Platform is not supported on the Safari browser. Users may experience unsupported behavior or be unable to access some console functionality when using Safari. | Use a supported browser, such as Google Chrome or Mozilla Firefox. |
| 25 | Compute | Iceberg table | Iceberg features might not be available on a cluster that has not been restarted after Iceberg support was enabled. | Restart the cluster before using Iceberg features. |
| 26 | Iceberg table | Procedure | The remove_orphan_files procedure is not supported.
|
|
| 27 | Iceberg table | Procedure | The rewrite_table_path procedure is not supported.
|
|
| 28 | Iceberg table | Procedure | The location argument of the snapshot procedure is not supported. | |
| 29 | Iceberg table | Query | Queries against the snapshots, history, metadata_log_entries, manifests, files, all_data_files, partitions, and refs metadata tables require a fully qualified four-part name when the Iceberg table is in the default catalog and schema. For example, SELECT * FROM default.<table_name>.snapshots is not supported.
|
Use the complete three-part names: <catalog>.<schema>.<table>.<metadata_table>. For example: SELECT * FROM default.default.<table_name>.snapshots.
|
| 30 | Iceberg table | Migrate | The drop_backup argument of the migrate procedure is not supported.
|
Run the migrate procedure without specifying the drop_backup argument.
|
| 31 | Iceberg table | Migrate | Migrating external tables to Iceberg is not supported because AI Data Platform does not permit overlapping table storage locations. | No current workaround. |
| 32 | Iceberg table | Migrate | Renaming an existing Iceberg table is not supported. | Restart the cluster and recreate the table with the required name. |