Eles querem ver você ficar calmo, metódico e isento de culpa sob pressão — restaure o serviço primeiro, diagnostique em segundo, e previna recorrência em terceiro. Use STAR.
Por que isso importa
text
INCIDENT ORDER
1. Stabilize — stop the bleeding (rollback, failover, mitigate)
2. Communicate — keep stakeholders updated on a clear channel
3. Diagnose — root cause once it's stable, not during
4. Prevent — a blameless post-mortem with action items
Exemplo prático
text
S: A deploy caused checkout errors for ~15% of users.
T: I was on call and had to restore service fast.
A: I rolled back the deploy first (service recovered in minutes), posted updates
every 10 minutes, then traced the cause to an unhandled null from a new API
field. I added a guard and a contract test.
R: Downtime stayed under 20 minutes. The post-mortem added the missing test to
CI so it can't recur.
