migration-safety-checker

This skill should be used when the user asks to "create a migration", "run makemigrations", "modify a model field", "rename a column", or mentions "schema change", "alter table", "database migration".

SKILL.md

GitHub

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?

Step 4: Generate Safe Migration

If user approves, create:

# Migration 1: Archive data
class Migration(migrations.Migration):
    dependencies = [('users', '0001_initial')]

    operations = [
        migrations.RunPython(archive_old_emails, reverse_code=restore_old_emails),
    ]

def archive_old_emails(apps, schema_editor):
    """Archive old_email data before dropping."""
    User = apps.get_model('users', 'User')
    EmailArchive = apps.get_model('users', 'EmailArchive')

    for user in User.objects.all():
        if user.old_email:
            EmailArchive.objects.create(
                user_id=user.id,
                email=user.old_email
            )

def restore_old_emails(apps, schema_editor):
    """Restore old_email data on rollback."""
    User = apps.get_model('users', 'User')
    EmailArchive = apps.get_model('users', 'EmailArchive')

    for archive in EmailArchive.objects.all():
        User.objects.filter(id=archive.user_id).update(
            old_email=archive.email
        )

# Migration 2: Drop column (run after deployment)
class Migration(migrations.Migration):
    dependencies = [('users', '0002_archive_emails')]

    operations = [
        migrations.RemoveField(
            model_name='user',
            name='old_email',
        ),
    ]

Safe Migration Patterns

Pattern 1: Renaming Field (3 Steps)

❌ UNSAFE:

operations = [
    migrations.RenameField('user', 'email', 'email_address'),  # ❌ Risky!
]

✅ SAFE (3-Step):

# Step 1: Add new field
operations = [
    migrations.AddField('user', 'email_address',
                       models.EmailField(null=True))
]

# Step 2: Data migration (copy old → new)
operations = [
    migrations.RunPython(copy_email_to_email_address)
]

# Step 3: Drop old field (separate deployment)
operations = [
    migrations.RemoveField('user', 'email')
]

Pattern 2: Changing Field Type (3 Steps)

❌ UNSAFE:

operations = [
    migrations.AlterField('product', 'price',
                         models.DecimalField(max_digits=10, decimal_places=2))
]

✅ SAFE:

# Step 1: Add new field with new type
operations = [
    migrations.AddField('product', 'price_decimal',
                       models.DecimalField(max_digits=10, decimal_places=2, null=True))
]

# Step 2: Data migration (convert & copy)
operations = [
    migrations.RunPython(convert_price_to_decimal)
]

# Step 3: Drop old field, rename new field
operations = [
    migrations.RemoveField('product', 'price'),
    migrations.RenameField('product', 'price_decimal', 'price')
]

Pattern 3: Adding Non-Nullable Field (2 Steps)

❌ UNSAFE:

operations = [
    migrations.AddField('user', 'organization',
                       models.ForeignKey('Organization'))  # ❌ No default!
]

✅ SAFE:

# Step 1: Add as nullable
operations = [
    migrations.AddField('user', 'organization',
                       models.ForeignKey('Organization', null=True))
]

# Step 2: Populate + make non-nullable
operations = [
    migrations.RunPython(assign_default_organizations),
    migrations.AlterField('user', 'organization',
                         models.ForeignKey('Organization', null=False))
]

Pattern 4: Adding Unique Constraint (Check First)

❌ UNSAFE:

operations = [
    migrations.AlterField('user', 'email',
                         models.EmailField(unique=True))  # ❌ Fails if duplicates!
]

✅ SAFE:

# Step 1: Check for duplicates
operations = [
    migrations.RunPython(check_and_fix_duplicate_emails)
]

# Step 2: Add unique constraint
operations = [
    migrations.AlterField('user', 'email',
                         models.EmailField(unique=True))
]

Production-Safe Practices

Use CONCURRENT Index Creation

from django.db import migrations
from django.contrib.postgres.operations import AddIndexConcurrently

class Migration(migrations.Migration):
    atomic = False  # Required for CONCURRENT operations

    operations = [
        AddIndexConcurrently(
            model_name='user',
            index=models.Index(fields=['email'], name='user_email_idx')
        )
    ]

Data Migration Template

def forward_migration(apps, schema_editor):
    """
    Safe data migration with batching.
    """
    Model = apps.get_model('app', 'Model')
    batch_size = 1000

    total = Model.objects.count()
    for offset in range(0, total, batch_size):
        batch = Model.objects.all()[offset:offset + batch_size]
        for obj in batch:
            # Transform data
            obj.new_field = transform(obj.old_field)
            obj.save(update_fields=['new_field'])

def reverse_migration(apps, schema_editor):
    """
    Reverse the migration.
    """
    Model = apps.get_model('app', 'Model')
    Model.objects.all().update(new_field=None)

class Migration(migrations.Migration):
    operations = [
        migrations.RunPython(forward_migration, reverse_migration)
    ]

Migration Testing Checklist

Before applying migration:

  • ✅ Run on copy of production database
  • ✅ Verify data integrity after migration
  • ✅ Test rollback works
  • ✅ Check migration duration (< 5 minutes ideal)
  • ✅ Review SQL with sqlmigrate
  • ✅ Confirm no downtime required
  • ✅ Have rollback plan ready

Validation Commands

Suggest running:

# Review SQL before applying
python manage.py sqlmigrate users 0002

# Check for issues
python manage.py makemigrations --check

# Dry run
python manage.py migrate --plan

# Apply to staging first
python manage.py migrate --database=staging

Integration with CI/CD

# .github/workflows/migrations.yml
- name: Check migrations safety
  run: |
    python manage.py makemigrations --check --dry-run --no-input
    python manage.py migrate --plan

Success Criteria

✅ No data loss migrations ✅ All migrations tested on staging ✅ Rollback plan exists ✅ Large table migrations use CONCURRENT ✅ Data migrations properly batched ✅ Reversible migrations

Behavior

Proactive enforcement:

  • Analyze migrations without being asked
  • Block unsafe migrations immediately
  • Suggest safe 3-step patterns
  • Generate data migration code
  • Explain WHY the migration is risky

Never:

  • Allow data loss migrations
  • Let users skip safety checks
  • Apply migrations without review
  • Just warn without blocking unsafe operations

When migration is unsafe:

  • Stop the process
  • Explain the risk
  • Provide safe alternative
  • Generate corrected migration code
  • User must acknowledge before proceeding