Flagship Docs

Appearance

Rollback Triggers ​

Safety is at the core of Flagship. Rollback Triggers ensure that if a new feature causes harm, it is instantly contained or completely reverted without requiring human intervention.

Automatic Rollbacks ​

Flagship continuously monitors your application's health using data from your Metrics Providers. You can define thresholds for critical metrics (e.g., Error Rate > 1%).

When an Automatic Rollback is triggered:

  1. Instant Reversion: The feature flag is immediately set back to its default value (usually false or off).
  2. Notification: Your team is alerted via Slack, PagerDuty, or other configured Alerting channels.
  3. Evidence: Flagship captures the exact metric state and timestamp of the rollback for post-incident Investigation.

Warning Triggers ​

Not every regression requires a full rollback. Some issues (e.g., slight latency increases or minor UI bugs) need attention but don't justify an immediate shutdown.

Warning Triggers halt the progression of a rollout.

  • Safety Lock: The rollout is frozen at its current percentage (e.g., if it was at 10%, it stays at 10%).
  • Prevention: The AI will not increase the percentage until a human reviews the situation and manually resumes the rollout.
  • Intervention: This gives your team time to investigate without exposing the new feature to any additional users.

Manual Rollbacks ​

Sometimes, the intuition of an engineer is faster than any metric. Every rollout includes a prominent Rollback button in the Flagship Dashboard.

A manual rollback:

  • Immediately kills the feature for 100% of users.
  • Halts any active automated plans.
  • Traces the action back to the user who initiated it for audit purposes.

IMPORTANT

To use Automatic Rollbacks, ensure you have connected a data source in the Onboarding section. Without live metrics, Flagship can only support manual rollbacks.