Databricks Migration Workshop: From Stored Procedures to Lakehouse (5-Day Intensive) Training Course

Introduction, Objectives, and Migration Strategy

• Course goals, alignment with participant profiles, and success criteria.
• High-level migration approaches and consideration of risks.
• Setup of workspaces, repositories, and lab datasets.

Day 1 — Migration Fundamentals and Architecture

• Lakehouse concepts, overview of Delta Lake, and Databricks architecture.
• Differences and implications for migration between SMP and MPP architectures.
• Medallion (Bronze→Silver→Gold) design and an overview of Unity Catalog.

Day 1 Lab — Translating a Stored Procedure

• Hands-on migration of a sample stored procedure to a notebook.
• Mapping temp tables and cursors to DataFrame transformations.
• Validation and comparison with the original output.

Day 2 — Advanced Delta Lake & Incremental Loading

• ACID transactions, commit logs, versioning, and time travel capabilities.
• Auto Loader, MERGE INTO patterns, upserts, and schema evolution.
• OPTIMIZE, VACUUM, Z-ORDER, partitioning, and storage tuning.

Day 2 Lab — Incremental Ingestion & Optimization

• Implementation of Auto Loader ingestion and MERGE workflows.
• Application of OPTIMIZE, Z-ORDER, and VACUUM; validation of results.
• Measurement of read/write performance improvements.

Day 3 — SQL in Databricks, Performance & Debugging

• Analytical SQL features: window functions, higher-order functions, and JSON/array handling.
• Reading the Spark UI, DAGs, shuffles, stages, tasks, and diagnosing bottlenecks.
• Query tuning patterns: broadcast joins, hints, caching, and spill reduction.

Day 3 Lab — SQL Refactoring & Performance Tuning

• Refactoring a heavy SQL process into optimized Spark SQL.
• Utilization of Spark UI traces to identify and resolve skew and shuffle issues.
• Benchmarking before/after results and documenting tuning steps.

Day 4 — Tactical PySpark: Replacing Procedural Logic

• Spark execution model: driver, executors, lazy evaluation, and partitioning strategies.
• Transforming loops and cursors into vectorized DataFrame operations.
• Modularization, UDFs/pandas UDFs, widgets, and reusable libraries.

Day 4 Lab — Refactoring Procedural Scripts

• Refactoring a procedural ETL script into modular PySpark notebooks.
• Introduction of parametrization, unit-style tests, and reusable functions.
• Code review and application of best-practice checklists.

Day 5 — Orchestration, End-to-End Pipeline & Best Practices

• Databricks Workflows: job design, task dependencies, triggers, and error handling.
• Designing incremental Medallion pipelines with quality rules and schema validation.
• Integration with Git (GitHub/Azure DevOps), CI, and testing strategies for PySpark logic.

Day 5 Lab — Build a Complete End-to-End Pipeline

• Assembly of a Bronze→Silver→Gold pipeline orchestrated with Workflows.
• Implementation of logging, auditing, retries, and automated validations.
• Execution of the full pipeline, validation of outputs, and preparation of deployment notes.

Operationalization, Governance, and Production Readiness

• Best practices for Unity Catalog governance, lineage, and access controls.
• Cost management, cluster sizing, autoscaling, and job concurrency patterns.
• Deployment checklists, rollback strategies, and runbook creation.

Final Review, Knowledge Transfer, and Next Steps

• Participant presentations of migration work and lessons learned.
• Gap analysis, recommended follow-up activities, and training materials handoff.
• References, further learning paths, and support options.