Migration methods for Teradata to Fabric Data Warehouse

Applies to: ✅ Warehouse in Microsoft Fabric

A Teradata migration has two primary steps: migrate metadata with Fabric Migration Assistant, then move data with Copy job, a Fabric pipeline, or COPY INTO. Complete the migration by validating results and rerouting connections.

For program planning, see Plan a migration from Teradata. For syntax and object mappings, see Translate Teradata SQL for Fabric Data Warehouse.

Stage Recommended method Result
Metadata Upload a zip of Teradata .sql and/or .bteq files to Migration Assistant Supported objects are translated and created in a new Warehouse
Remediation Review Objects to fix and use documented mappings or Copilot suggestions Required objects are corrected and deployed
Data Use Copy job for a guided transfer; use staged files and COPY INTO for large bulk loads Source data is loaded and reconciled
Cutover Validate workloads and reroute loading and reporting connections Applications use Fabric Data Warehouse

Note

Migration Assistant translates and deploys metadata and code. Data movement runs as a separate Copy job or ingestion step in the migration workflow.

Prerequisites

Before you begin, prepare:

  • A Fabric workspace with active capacity.
  • A zip archive of .sql and/or .bteq files containing Teradata table, view, procedure, function, macro, and other required object definitions.
  • All dependencies for the schemas in the migration wave.
  • A destination Warehouse name and collation choice.
  • Source credentials if you plan to use Copy job for data movement.

Include DDL definitions rather than files that contain only standalone SELECT statements. Migration Assistant ignores query-only files that don't define an object.

Migrate metadata with Migration Assistant

Migration Assistant supports Teradata as a source and translates uploaded Teradata metadata to Fabric-compatible T-SQL.

  1. In your Fabric workspace, select Migrate.
  2. Under Migrate to a warehouse, select the Teradata (Preview) source tile.
  3. On Set the source, select Choose file, browse your zip file, and then select Next.
  4. Review the detected source objects and select the schemas or individual objects for this migration wave.
    • When you select an individual object, you can also see the dependent objects selected for migration.
  5. On Set the destination, select the workspace, enter the new Warehouse name, and choose the collation.
  6. Review the inputs, then select Migrate.
  7. Wait while Migration Assistant translates supported metadata and creates the new Warehouse.
  8. Review the metadata migration summary. Expand Show migrated objects to see each object and its state.
  9. Select an object to review datatype or SQL adjustments in Details.
  10. Export the summary when you need an offline record for review or migration tracking.

Objects with missing dependencies or unsupported constructs appear under Objects to fix.

Fix metadata translation problems

Use Migration Assistant to review and correct objects that it doesn't create automatically:

  1. Open Fix problems and select an object.
  2. Review the source definition, automatic translation comments, and error details in the shared query.
  3. If Copilot is enabled for your tenant and capacity, select Fix query errors to generate a suggested correction.
  4. Review all AI-generated changes. Adjust the script as needed.
  5. Select Run to validate the script and create the object.
  6. Repeat for each required object. Skip objects that aren't needed in the target.

Copilot can make mistakes. Compare corrected code with Teradata behavior and test boundary, null, precision, and error cases.

Copy data

After the required table metadata exists, choose one data movement method.

Use Copy job

Use Copy job for a guided full or incremental transfer:

  1. In Migration Assistant, open Copy data, and then select Use a copy job.
  2. Name the job and connect to Teradata by using the source credentials.
  3. Select the tables to copy.
  4. Select the Warehouse from the OneLake catalog.
  5. Review table and column mappings.
  6. Choose a one-time full copy for the initial migration, or incremental copy for a coexistence period when the source configuration supports it.
  7. Select Save + Run, and then monitor row counts, duration, and failures.

Use staged files and COPY INTO

For large historical transfers, bulk unload Teradata tables to ADLS Gen2. Use Parquet files when supported, keep an extraction manifest, and load the files by using COPY INTO.

Example:

COPY INTO dbo.FactSales
FROM 'https://<storage-account>.dfs.core.windows.net/<container>/factsales/'
WITH (
    FILE_TYPE = 'PARQUET'
);

Use trusted workspace access and a workspace identity when the storage account network configuration requires them. Use a Fabric pipeline instead when the flow requires multi-step orchestration, transformations, branching, or custom recovery.

Replace Teradata loading utilities

Teradata utilities represent both data movement and orchestration behavior. Recreate the complete behavior rather than translating only the load statement.

Teradata utility Fabric method Migration pattern
FastLoad COPY INTO or Copy job Bulk unload to staged files, create an empty target table, and load in parallel
MultiLoad COPY INTO plus MERGE Load files into a staging table, validate, then merge inserts, updates, and deletes into the target
Teradata Parallel Transporter Fabric pipeline with copy and script activities Recreate operators, sequencing, error handling, and dependencies as explicit pipeline steps
BTEQ .IMPORT COPY INTO a table Replace client-side import behavior with a governed staged-file load
BTEQ .EXPORT Pipeline, notebook, or supported export pattern Select the destination and file format based on the consuming system

Validate and reroute connections

Before cutover:

  • Reconcile row counts, aggregates, nulls, decimals, timestamps, and collation-sensitive values.
  • Test views, procedures, functions, BTEQ replacements, reports, and semantic models.
  • Test failed-load replay, partial failure, incremental synchronization, and peak concurrency.
  • Compare source and target results for representative business cycles.
  • Run a final delta load and complete the cutover checklist.
  • Update reporting, semantic model, application, and ETL/ELT connections to the Fabric Data Warehouse.

Related content

  • Last updated on