Migration
Azure Synapse to Databricks migration: dedicated pools, Spark and pipelines
How to migrate Azure Synapse Analytics to Azure Databricks: dedicated SQL pools to Delta and Databricks SQL, T-SQL conversion, Synapse Spark notebooks, serverless views over ADLS, pipelines and Power BI.
Hemant Walchale
· 2 min read
Azure Synapse customers now have two natural successors: Microsoft Fabric and Azure Databricks. Databricks is the common choice for organizations where data engineering, machine learning and AI matter as much as reporting, or that want the same platform across clouds. Each Synapse component has a clear destination.
This is the approach we recommend.
1. Inventory
List dedicated SQL pool objects (tables, views, procedures), serverless SQL pool views, Synapse pipelines and data flows, Spark pools with notebooks and libraries, linked services, triggers and Power BI datasets. Run history shows what is used; retire the rest.
2. Map Synapse to Databricks
| Azure Synapse | On Azure Databricks |
|---|---|
| Dedicated SQL pool tables | Delta tables in Unity Catalog |
DISTRIBUTION (HASH, ROUND_ROBIN, REPLICATE), columnstore indexes |
Not needed; liquid clustering for large filtered tables |
| Dedicated pool stored procedures (T-SQL) | Databricks SQL or Python; rewrite procedural logic |
| Serverless SQL views over ADLS | Unity Catalog external tables or Delta tables on the same storage |
| Spark pools and notebooks | Databricks notebooks and jobs (mssparkutils → dbutils) |
| Synapse pipelines and data flows | Lakeflow Jobs and Lakeflow Declarative Pipelines (or keep Azure Data Factory for orchestration) |
| Linked services and credentials | Unity Catalog connections, storage credentials and secret scopes |
| Triggers | Job schedules and file-arrival triggers |
| Power BI on Synapse | Power BI on Databricks SQL warehouses |
3. Convert T-SQL
Dedicated pool T-SQL converts to Spark SQL with common changes: TOP to LIMIT, ISNULL to COALESCE, + string concatenation to concat() or ||, CTAS distribution options removed, temporary tables to temporary views, and TRY...CATCH procedural logic rewritten. Test date functions and implicit conversions.
4. Reconcile, repoint and pause pools
Compare every migrated table by domain, repoint Power BI datasets to Databricks SQL warehouses once owners sign off, then pause and delete dedicated SQL pools and Spark pools as domains move.
The short version
Register existing lake data in Unity Catalog instead of copying it, load dedicated pool tables as Delta, drop distribution settings, convert T-SQL, port Spark notebooks, rebuild pipelines as Lakeflow Jobs, reconcile, repoint Power BI and pause pools by domain.
Compare with Synapse to Microsoft Fabric and Databricks vs Microsoft Fabric, or browse all migration guides.
Not sure where your own platform stands? Take the free 3-minute data platform scorecard, or see how we work on Databricks. Unfamiliar terms are explained in the data engineering glossary.