Supported sources
Every source track in the Databricks M&M program is covered — Microsoft SQL, Snowflake, AWS Redshift, Azure Synapse, Oracle, Hadoop, and Teradata — plus Netezza beyond the list.
Airlift covers all seven source tracks in the Databricks Migration & Modernization
program, plus Netezza. The matrix is code, not prose: every row below is an entry in
SOURCE_SYSTEM_PROFILES (@fabricorg/airlift), and the repository contract test fails
if any program track loses coverage.
| Source | M&M program track | Deterministic transpile | Analyzer surfaces | Profiler | Reconcile |
|---|---|---|---|---|---|
| SQL Server | Microsoft SQL Migration | mssql | MS SQL Server, SSIS, SSRS | ✓ | ✓ |
| Snowflake | Snowflake Migration | snowflake | Snowflake | ✓ | ✓ |
| AWS Redshift | AWS Redshift Migration | redshift | Redshift | ✓ | ✓ |
| Azure Synapse | Azure Synapse Migration | synapse | Synapse, ADF | ✓ | via SQL Server connector / parity suites |
| Oracle | Oracle Migration | oracle | Oracle, ODI | ✓ | ✓ |
| Hadoop | Hadoop Migration | agent lane | Hive, Impala, Oozie, Pig, Sqoop | — | parity suites |
| Teradata | Teradata Migration | teradata | Teradata | pending verification | ✓ |
| Netezza | — (beyond program) | netezza | Netezza | — | parity suites |
Column meanings, mapped to the Lakebridge contract:
- Deterministic transpile — the Lakebridge
--source-dialectfor the first-pass conversion. "Agent lane" means the residue path (Switch/repair agent with sandboxed execution) carries conversion; the funnel, certificates, and cutover gates are identical either way. - Analyzer surfaces — the
--source-techvalues the assessment sweeps, including the ETL/orchestration/report adjuncts (SSIS, ADF, ODI, Oozie) that real estates contain. - Profiler — live-source sizing/complexity from the Lakebridge profiler.
- Reconcile — the Lakebridge reconcile module as a parity-certificate evidence source; where it lacks a connector, the databricks-testkit parity suites (Phase 2) produce the evidence instead.
One estate model, per-source depth
The estate/object/wave ontology, the parity-certificate schema, the cutover gates, and the evidence export are source-agnostic — a Teradata migration and a Synapse migration produce the same funnel, the same certificate depths, and the same audit trail. What varies per source is only the assessment surfaces and the deterministic conversion coverage, which is exactly what the matrix records. This is the multi-source-consistency moat: one governed method across every specialization badge.
Quickstart
Run the Airlift walking skeleton locally — the seeded console, the governed pipeline tests, and the mock cutover worker — with no Databricks workspace required.
The migration method
Assess, plan, convert, certify, cut over, attest — six governed stages, the residue tripwire, and why the factory beats the toolkit.