Diagnostic Questions¶
Before revealing the investigation path:¶
-
The canary deployment shows 0% error rate and sub-100ms latency, but users report seeing wrong prices. Standard canary metrics look healthy. What category of failure would NOT be detected by error rate and latency metrics?
-
The canary pod returns the correct price when called directly via
curl localhost:8080. But users on the canary see wholesale prices. What could cause a difference between what the pod serves directly and what users see? -
The canary pod's IP appears in the endpoints for both
product-search-canaryandwholesale-pricing-api. How did this happen, and what Kubernetes mechanism causes a pod to appear in a Service's endpoints? -
The canary deployment has
tier: pricinginstead oftier: frontend— a copy-paste error from a different deployment template. Why is this a Kubernetes label issue (Domain C) rather than a DevOps deployment process issue (Domain A)? -
How would you detect endpoint contamination (a pod appearing in an unintended Service) before it affects users? What automated check or admission controller would catch this?