Overview

Instructions

Step 1: Discovery and Assessment

Inventory all source tables with metadata for migration planning.

from pyspark.sql import SparkSession
from dataclasses import dataclass

spark = SparkSession.builder.getOrCreate()

@dataclass
class TableInventory:
    database: str
    table: str
    table_type: str
    format: str
    row_count: int
    size_mb: float
    columns: int
    partitions: list[str]

def assess_hive_metastore() -> list[TableInventory]:
    """Inventory all Hive Metastore tables for migration planning."""
    inventory = []
    databases = [r.databaseName for r in spark.sql("SHOW DATABASES").collect()]

    for db in databases:
        tables = spark.sql(f"SHOW TABLES IN hive_metastore.{db}").collect()
        for t in tables:
            table_name = f"hive_metastore.{db}.{t.tableName}"
            try:
                detail = spark.sql(f"DESCRIBE DETAIL {table_name}").first()
                schema = spark.table(table_name).schema

                inventory.append(TableInventory(
                    database=db,
                    table=t.tableName,
                    table_type=detail.format or "unknown",
                    format=detail.format or "unknown",
                    row_count=spark.table(table_name).count(),
                    size_mb=detail.sizeInBytes / 1048576 if detail.sizeInBytes else 0,
                    columns=len(schema),
                    partitions=detail.partitionColumns or [],
                ))
            except Exception as e:
                print(f"  Skipping {table_name}: {e}")

    return inventory

# Generate migration plan
tables = assess_hive_metastore()
tables.sort(key=lambda t: t.size_mb, reverse=True)

print(f"\nTotal tables: {len(tables)}")
print(f"Total size: {sum(t.size_mb for t in tables):.0f} MB")
print(f"\nTop 10 by size:")
for t in tables[:10]:
    print(f"  {t.database}.{t.table}: {t.size_mb:.0f}MB, {t.row_count:,} rows, {t.format}")

Step 2: Schema Migration

# Schema conversion for common type mismatches
TYPE_MAP = {
    # Hadoop/Hive types → Delta Lake/Spark types
    "CHAR": "STRING",
    "VARCHAR": "STRING",
    "TINYINT": "INT",
    "SMALLINT": "INT",
    "BINARY": "BINARY",
    # Snowflake types
    "NUMBER": "DECIMAL",
    "VARIANT": "STRING",  # Store as JSON string, parse in Silver
    "TIMESTAMP_NTZ": "TIMESTAMP",
    "TIMESTAMP_TZ": "TIMESTAMP",
    # Redshift types
    "SUPER": "STRING",
    "TIMETZ": "TIMESTAMP",
}

def generate_create_table(source_table: str, target_table: str) -> str:
    """Generate CREATE TABLE DDL with type conversions."""
    schema = spark.table(source_table).schema
    cols = []
    for field in schema:
        dtype = TYPE_MAP.get(str(field.dataType).upper(), str(field.dataType))
        cols.append(f"  {field.name} {dtype}")

    return f"""CREATE TABLE IF NOT EXISTS {target_table} (
{',\n'.join(cols)}
) USING DELTA
TBLPROPERTIES (
    'delta.autoOptimize.optimizeWrite' = 'true',
    'delta.autoOptimize.autoCompact' = 'true'
);"""

Step 3: Data Migration with Validation

def migrate_table(
    source_table: str,
    target_table: str,
    method: str = "ctas",
    batch_size_mb: int = 500,
) -> dict:
    """Migrate a table with validation."""
    result = {"source": source_table, "target": target_table, "method": method}

    if method == "sync":
        # In-place metadata migration (fastest, no data copy)
        spark.sql(f"SYNC TABLE {target_table} FROM {source_table}")

    elif method == "deep_clone":
        # Delta-to-Delta with history preservation
        spark.sql(f"CREATE TABLE {target_table} DEEP CLONE {source_table}")

    elif method == "ctas":
        # Full data copy (works with any source format)
        source_size_mb = spark.sql(
            f"DESCRIBE DETAIL {source_table}"
        ).first().sizeInBytes / 1048576

        if source_size_mb > batch_size_mb:
            # Batch large tables by partition or row number
            spark.sql(f"""
                CREATE TABLE {target_table}
                USING DELTA
                AS SELECT * FROM {source_table}
            """)
        else:
            spark.sql(f"CREATE TABLE {target_table} AS SELECT * FROM {source_table}")

    elif method == "jdbc":
        # External database migration
        df = (spark.read
            .format("jdbc")
            .option("url", f"jdbc:postgresql://host:5432/db")
            .option("dbtable", source_table)
            .option("fetchsize", "10000")
            .load())
        df.write.format("delta").saveAsTable(target_table)

    # Validate
    src_count = spark.table(source_table).count()
    tgt_count = spark.table(target_table).count()
    result["source_rows"] = src_count
    result["target_rows"] = tgt_count
    result["match"] = src_count == tgt_count
    result["status"] = "OK" if result["match"] else "MISMATCH"

    return result

# Migrate with validation
result = migrate_table(
    "hive_metastore.legacy.customers",
    "analytics.migrated.customers",
    method="ctas",
)
print(f"{result['source']} -> {result['target']}: "
      f"{result['source_rows']:,} rows [{result['status']}]")

Step 4: Snowflake / Redshift Migration

# Snowflake: Use Lakehouse Federation or Unload + Auto Loader
# Option A: Lakehouse Federation (query in place, no copy)
spark.sql("""
    CREATE FOREIGN CATALOG snowflake_catalog
    USING CONNECTION snowflake_conn
    OPTIONS (database 'PROD_DB')
""")
# Query directly: SELECT * FROM snowflake_catalog.schema.table

# Option B: Unload to S3 + ingest
# In Snowflake:
# COPY INTO @my_s3_stage/export/customers/
# FROM PROD_DB.PUBLIC.CUSTOMERS
# FILE_FORMAT = (TYPE = PARQUET);

# In Databricks:
df = spark.read.parquet("s3://migration-bucket/export/customers/")
df.write.format("delta").saveAsTable("analytics.migrated.customers")

# Redshift: Unload to S3 + Auto Loader
# In Redshift:
# UNLOAD ('SELECT * FROM prod.customers')
# TO 's3://migration-bucket/redshift/customers/'
# FORMAT PARQUET;

# In Databricks:
(spark.readStream
    .format("cloudFiles")
    .option("cloudFiles.format", "parquet")
    .option("cloudFiles.schemaLocation", "/checkpoints/migration/schema")
    .load("s3://migration-bucket/redshift/customers/")
    .writeStream
    .format("delta")
    .option("checkpointLocation", "/checkpoints/migration/data")
    .toTable("analytics.migrated.customers"))

Step 5: ETL Pipeline Conversion

# Convert Oozie/Airflow jobs to Databricks Asset Bundles
# Before (Oozie/spark-submit):
#   spark-submit --class com.company.ETL --master yarn app.jar
#   hive -e "INSERT OVERWRITE TABLE target SELECT * FROM staging"

# After (Asset Bundle):
# databricks.yml resources:
"""