The following lists the fixed and known issues in this release. If you experience any of the following issues, open an incident with Teradata Customer Support and include the Key in your description.
Known Issues
| Key | Description |
|---|---|
| VEL - 2559 | When Vector store is created with non-English data, we may see accuracy issue when using ask or similarity search API on such vector store. Workaround: None. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on AWS |
User Defined Functions
| Key | Description |
|---|---|
| UDF - 1709 | BYOM functions may fail after a database restart. This is due to a problem with UDF containers being out of sync. Workaround: None. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on AWS |
| UDF - 1427 | It was discovered that there are issues in supporting more than one third party UDF solutions, including the following:
Workaround: There is no workaround to install more than one third party UDF solutions until this issue is resolved. If running into any issues, open a case from Support Portal to engage Teradata Support for assistance.
Deployments: AWS is the only platform on Lake supporting more than one third-party solution. |
| Key | Description |
|---|---|
| TDGSS-11273 | The Power BI job failed after the TLS connection, established through the .NET data provider, was unexpectedly disconnected during execution. Workaround: Add proper serialization of TLS read/write operations to prevent connection instability during Power BI jobs. Deployments: Lake on Azure| GC| AWS |
| Key | Description |
|---|---|
| SQLES-14124 | If encryption is enabled for internal bucket, it can cause failure of provisioning of the QueryGrid component. Workaround: None. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on Azure| GC| AWS |
Open Table Format
| Key | Description |
|---|---|
| OTF-3749 | Query failures observed due to OTF Java engine hitting the Out Of Memory exceptions. Workaround: The memory issue occurs when the OTF engine is subjected to a high volume of concurrent queries over an extended period. While typically short-lived, in rare cases this can cause the Java-based OTF engine to enter a stale state, disrupting communication with AMP processes. When this happens, restarting the Java engine usually resolves the problem. To initiate a restart, open a support ticket. Deployments: Lake on Azure| GC| AWS |
| OTF-3516 | Failure during synchronization of Iceberg metadata in Databricks Unity catalog using Spark SQL. Workaround: The issue happens when a spark job is submitted on the Databricks Spark cluster to perform Iceberg metadata sync operation. This is an intermittent issue and the current workaround is to rerun the query. Deployments:Databricks Unity/Iceberg write operations ONLY on all CSPs. |
| Key | Description |
|---|---|
| HARM-6898 | The incorrect error is returned to the user causing confusion. Workaround: None. For now all 4500 memory errors should be viewed as an LSN not found error, which means the system was restarted and no reconnect is allowed. Deployments: Lake on Azure| GC| AWS |
| Key | Description |
|---|---|
| CCP-11041 | If user excludes some object from a database having more than 10K objects then restore job will not restore objects beyond 10k objects from that database. Workaround: Do not exclude tables from a database when it has more than 10k tables or restore the complete database. Deployments: Lake on AWS |
| Key | Description |
|---|---|
| BYOA-3179 | A Timeout Error is observed when org admin user activates Anaconda feature from the VantageCloud Lake Console. Workaround: None. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on GC |
Fixed Issues
| Key | Description |
|---|---|
| VMO - 1832 | The evaluation job for BYOM (except Python/R) with custom metrics fails. Workaround: Cannot evaluate BYOM (except Python/R) with custom metrics but can be evaluated using default metrics. While importing the default metrics can be selected for monitoring. Deployments: Lake on Azure| GC| AWS |
| VMO - 1827 | The evaluation job fails for BYOM DataRobot with error - ValueError: Classification metrics cannot handle a mix of binary and unknown targets. Workaround: None. Cannot evaluate BYOM DataRobot. Deployments: Lake on Azure| GC| AWS |
| VMO - 1814 | While running a compute statistics job for BYOM model in Demo project (pre-loaded ModelOps project), it fails with error - categorical_features = f for f in feature_names if feature_summaryf.lower()] == 'categorical'] KeyError: 'numtimesprg' Workaround: Change database to td_modelops in the dataset template and for all (dataset template and datasets) SQL queries add the database as td_modelops. (Only for the Demo project) Example - Change SELECT * FROM pima_patient_features to SELECT * FROM td_modelops.pima_patient_features Deployments: Lake on Azure| GC| AWS |
| VMO-1747 | ModelOps provisioning could fail due to Private DNS timeout after 5 minutes. Workaround: Delete ModelOps and retry ModelOps provisioning. Deployments: Lake on AWS |
| VMO-1716 | While importing a BYOM and generating a prediction expression, the expressions do not list out and the loading icon shows up. The error message can be - 'Cannot convert undefined or null to object'. Workaround: Instead of generating a prediction expression, the user can manually enter the prediction expression. Example - CAST(CAST(json_report AS JSON).JSONExtractValue('$.predicted_HasDiabetes') AS INT). The prediction expression cannot be validated Deployments: Lake on Azure| GC| AWS |
| Key | Description |
|---|---|
| TCOPTT-1012 | In VantageCloud Lake, when upgrading to the August 2024 release or later, if the PDCR history table PDCRDATA.AcctgDtl_Hst, Acctg_Hst, MonitorSession_Hst, TDWMThrottleStats_Hst or TDWMUtilityStats_Hst contains data (e.g., from a previous migration), then it will not be converted to an OFS table and the table's corresponding collection job will not run. Workaround: Run SQL statements to rename the existing table and create a new empty table in OFS. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on AWS |
| Key | Description |
|---|---|
| OTF-3207 | The issue was seen when a Teradata Stored Procedure (TDSP) is created prior to 20.00.25.14 and is then called on a later release. In this case, Failure 7551 may return. The issue could also be seen when a table with Partitioned Primary Index (PPI) is created prior to 20.00.25.14 and is then used in a SELECT statement on a later release. Workaround: For a TDSP, recompile the TDSP before calling it. For a PPI table, revalidate the table before using it in a DML. Deployments: Lake on Azure| GC| AWS |
| OTF-3171 | Intermittent failures with mixed workload on AWS OTF queries. Workaround: Restart Java OTF UDF Server. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on AWS |
| OTF-3023 | Intermittent issue where OTF query cannot open the input stream. Workaround: Restart Java OTF UDF Server. Submit a case from Support Portal to engage Teradata Support for assistance. Deployments: Lake on AWS |
| Key | Description |
|---|---|
| HARM-6749 | Some data connections could fail causing a load job to hang for long periods before failing. Now load sessions are more likely to succeed and if they still fail they will fail immediately. |
| Key | Description |
|---|---|
| DINSIGHTS-1273 | ASK API requests may intermittently fail with a 400 Bad Request due to an underlying 429: Rate limit is exceeded error from the Azure OpenAI service. Workaround: Retry the request after a short delay (e.g., 1 second). Deployments: Lake on Azure |