Nova fixes a crash-looping deploy
Rather not read it all? Ask Nova this. It walks you through each step and does the ones it can when you approve.
A real incident, INC-79, from the first read to the postmortem.
pods-probe-web, an app Bob's team owns, restarts in a loop. An investigation opens at 23:51 PDT, and Dana, the on-call engineer, picks it up.
1. Nova finds the cause
Nova reads the app's status, events, deploy history and configuration, asks before reading the logs (Dana allows it at 23:52), tests 4 possible causes and rules out 3.
Cause found: "The container runs as a non-root user and cannot write to a path nginx requires at startup", proved by the chown error it quotes from the app's logs.
2. Dana approves the fix
Redeploy pods-probe-web with the nginx-unprivileged image, low risk and reversible. Dana selects Let Nova fix it at 23:55.
3. Nova checks the app
"pods-probe-web recovered: RUNNING, no warnings in the last 5 min". Bob's team has its app back.
4. Sam reads the postmortem
Result: a 17-minute investigation, a fix a person approved, a verified recovery and a postmortem from the record. Every step in detail: Investigate a failing app.