Clone Production

Clone production database

Clone a production database with automatic PII anonymization.

Usage

# Configure anonymization rules
xata clone config --source-url <production-connection-string>

# Run the clone
xata clone start --source-url <production-connection-string>

Example workflow

# 1. Set up your production connection string
export PROD_URL="postgres://user:pass@prod-host:5432/mydb"

# 2. Configure what gets anonymized (interactive)
xata clone config --source-url $PROD_URL --mode=prompt

# 3. Run the clone
xata clone start --source-url $PROD_URL

# 4. Connect to your anonymized copy
psql $(xata branch url)

Auto-detected PII columns

Xata automatically detects and anonymizes columns named:

• email → fake emails
• name, first_name, last_name → fake names
• phone → fake phone numbers
• ssn → masked SSN
• address → fake addresses
• And more...

Custom transforms

Edit .xata/clone.yaml to customize:

tables:
  users:
    columns:
      email:
        transform: email
      custom_field:
        transform: redact

Available transforms

• email — Generate fake email
• name — Generate fake name
• phone — Generate fake phone
• address — Generate fake address
• redact — Replace with [REDACTED]
• hash — One-way hash (preserves uniqueness)
• null — Set to NULL

Why this matters for AI apps

• Debug with real data distributions
• Test edge cases only production reveals
• Maintain referential integrity across tables
• Stay compliant with privacy regulations

Tips

• Use a read-replica connection string if available
• Avoid pooler/pgbouncer endpoints (clone uses long transactions)
• Review .xata/clone.yaml before running in CI/CD