In the 2 pictures below I am comparing the metric from K8s apiserver_client_certificate_expiration_seconds_count{job="apiserver"} for my 3 master nodes on K8s 1.14.3. A single day should have 86400 seconds, how come at the start of the day for a given node Prometheus displays 59565 at 00:00 and at 24:00 it shows 205534 seconds? Does that mean that the metric reported by k8s is broken or my query on the graphical dashboard has something in it that I don't understand? In theory it should mean that many seconds have expired from the api server cert, but the math doesn't match the specifications? enter image description here

enter image description here


Based on community answer to that question it turned out there isn't count of the seconds, it is count of all samples for given histogram. That means I had ~146,000 HTTP connections to that API server in that period.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.