Clock skew JWT validation
Your task
Your auth-gated API suddenly returns 401s for ~8% of requests, but only from one of three Kubernetes node pools, and only in bursts. The 401s come with 'token not yet valid' (nbf) and occasionally 'token expired' errors validating short-lived (60s) JWTs. The affected pods were rescheduled onto a newly added node group two days ago. No code or key-rotation change. Mean latency and CPU are normal. How do you triage and what's the fix?
It covers
- Clock skew
- Tls
- Jwt expiry