Incident Runbooks — Common Scenarios

401 storm on /api/**

• Symptom: sharp rise in 401s, SPAs stuck on login
• Checks: cookie domain/scope, /auth/session truth, CORS allowlist; ForwardAuth headers present
• Action: fix cookie domain or allowlist; redeploy; verify 200s resume

PDP degraded

• Symptom: 5xx/latency in PDP client logs; 403s increase
• Checks: PDP health, network path, PDP caches
• Action: fail open only if policy allows; otherwise communicate outage; consider increasing deny TTLs temporarily

IdP down

• Symptom: login/callback failures; refresh errors
• Checks: IdP health; retry/backoff metrics; edge routes
• Action: keep existing sessions; disable features requiring fresh tokens; escalate

Redis unavailable

• Symptom: session lookups fail; 401s
• Checks: Redis health; persistence; memory policy
• Action: restore Redis; consider rolling session TTL; communicate re‑login requirement