migration-safety-checker

Migration Safety Checker

Validates Django migrations are production-safe before they cause data loss.

Activation Triggers

This skill activates when:

• Running makemigrations
• Modifying model fields
• Mentioning "migration", "schema", "alter table"
• Adding/removing/renaming fields
• Changing field types
• Creating or modifying migration files

Safety Requirements

All migrations MUST be:

• ✅ Reversible - Can roll back safely
• ✅ No data loss - Preserve existing data
• ✅ Tested - Run on staging first
• ✅ Safe for production - No downtime operations
• ✅ Documented - Explain what changes and why

High-Risk Operations

These operations require special attention:

🔴 CRITICAL RISK - Data Loss Possible

1. Dropping columns without data migration
2. Renaming fields without data migration
3. Changing field types (data conversion issues)
4. Adding non-nullable fields without default
5. Removing tables with data

🟡 MEDIUM RISK - Requires Careful Planning

1. Adding indexes on large tables (use CONCURRENTLY)
2. Adding unique constraints (check for duplicates first)
3. Changing max_length (data truncation risk)

🟢 LOW RISK - Generally Safe

1. Adding nullable fields
2. Adding fields with defaults
3. Creating new tables
4. Adding non-unique indexes

Validation Process

Step 1: Analyze Migration File

When migration is created:

# Generated migration
class Migration(migrations.Migration):
    operations = [
        migrations.RemoveField(
            model_name='user',
            name='old_email',  # 🔴 DATA LOSS RISK!
        ),
    ]

Step 2: Identify Risks

Detect:

• Operation type: RemoveField
• Field: old_email
• Risk: 🔴 CRITICAL - Dropping column loses data!

Step 3: Block Unsafe Migration

⛔ UNSAFE MIGRATION DETECTED

Operation: RemoveField(model_name='user', name='old_email')

Risk: 🔴 CRITICAL - Data loss!

Problem:

• Drops old_email column immediately
• All existing email data will be PERMANENTLY LOST
• Cannot be rolled back after data is gone

Safe Approach: Use 3-step migration pattern:

1. Data migration: Archive old_email data
2. Wait for deployment
3. Drop column in separate migration

Should I generate the safe migration sequence?