Automatic rollbacks

Auto rollbacks are a safety feature that rollback deployments when new, high volume anomalies or errors are detected via Sentry, Datadog, or other monitoring tools.

When enabled this helps maintain system stability by reverting to the default branch when critical issues are detected during or after deployment.

In your project settings you’ll find a section for enabling and configuring automatic rollbacks:

You can configure:

• Target environment
• Detection mode
• Error threshold (minimum: 1)

Detection modes

Code-related only

• Triggers on code-specific issues (e.g., new errors in logs)
• Ignores deployment process issues
• Best for: Teams focused on code quality

Deployment-time only

• Triggers on deployment process issues
• Catches infrastructure errors and service problems
• Ignores post-deployment code issues
• Best for: Critical deployment stability

Code OR deployment issues

• Most sensitive mode
• Triggers if either type of issue occurs
• Best for: Maximum protection against failures

Code AND deployment issues

• Most conservative mode
• Requires both types of issues to trigger
• Best for: Minimizing false positives

How it works

When a rollback is triggered:

1. The current deployment is stopped
2. A new deployment starts for the default branch (e.g. main)
3. The team receives a notification

Best practices

A few tips for initial setup of automatic rollbacks:

• Automatic rollbacks rely on a stable, safe default branch — if you’re not sure if main is safe, rollbacks won’t work well
• Start with higher error thresholds
• Configure team notifications

Limitations

• Only rolls back to default branch
• Minimum threshold: 1 error
• Requires error correlation setup
• No rollback to specific versions