dbt-migration_skill

Overview

This skill provides a complete, repeatable workflow for migrating tables, views, and stored procedures into production-quality dbt projects on Snowflake. It orchestrates discovery, planning, placeholder creation, view/procedure conversion, testing, and deployment while delegating platform-specific syntax translation to source-specific conversion helpers. The goal is predictable migrations with clear validation gates and a safe cutover path.

How this skill works

The workflow runs seven sequential phases: Discovery, Planning, Placeholders, Views, Table Logic, Testing, and Deployment. It first inventories and assesses source objects, creates contract-first placeholder models, converts views and procedural ETL into declarative dbt models, then validates results against the source using checksums, counts, and business-rule tests before cutover. Platform-specific SQL translation (e.g., Oracle→Snowflake) is handled by dedicated conversion utilities before dbt conversion.

When to use it

• Migrating an on-prem or cloud RDBMS (SQL Server, Oracle, Teradata, etc.) into dbt on Snowflake
• Replatforming legacy stored procedures and ETL jobs into declarative dbt models
• Standardizing naming, layering, and materializations across a newly migrated warehouse
• Validating parity between source systems and migrated dbt outputs before cutover
• Creating a repeatable process for large-scale, multi-object migrations

Best practices

• Start with a full inventory and dependency graph to define safe migration order
• Create placeholder models with explicit datatypes and where false before adding logic
• Convert simple views first to build confidence, then tackle complex views and procedures
• Map procedural ETL patterns to dbt materializations (incremental, table, snapshots)
• Automate validation: row counts, checksums, aggregates, and business-rule tests
• Define a clear cutover and rollback plan with post-deployment monitoring

Example use cases

• Use SnowConvert AI to translate Oracle procedures to Snowflake, then convert to dbt models
• Break a monolithic SCD Type 2 procedure into snapshot + incremental dbt models with tests
• Create placeholders for 1,000+ target tables to enable parallel development and downstream testing
• Run checksum and aggregate validations to prove parity before switching BI to Snowflake
• Implement scheduled production runs with Snowflake tasks or dbt Cloud after cutover