In template Kubernetes nodes by HTTP total requests/limits for a given node are calculated by summing up requests/limits of all pods present on it. However the template does not account the fact that only Running/Pending pods should be included into the calculation - and includes pods which are Completed or Failed.

Steps to reproduce:

1. Have a kubernetes cluster observed by the Kubernetes nodes by HTTP template
2. Have a node which has pods in a Completed and/or Failed states (for example, letfovers from cronjobs)
3. See that requests/limits values reported on Zabbix are larger than actual values.

Proposed solution:

Update preprocessing step on Requests/Limits from:

$.Pods[?(@.hostname == "{#NAME}")].containers.<requests/limits>.<cpu/memory>.sum() 

to

$.Pods[?(@.hostname == "{#NAME}" && (@.phase == "Running" || @.phase == "Pending"))].containers.<requests/limits>.<cpu/memory>.sum()  

on the 

kube.node.<requests/limits>.<cpu/memory>[{#NAME}] 

item prototypes