CPU & Memoria per Pod/Container

Filtro standard namespace applicativi da applicare ovunque:

namespace!~"kube-.*|openshift-.*"

CPU

CPU usage per pod (somma di tutti i container)

sum by (namespace, pod) (
  rate(container_cpu_usage_seconds_total{
    namespace!~"kube-.*|openshift-.*",
    container!="", container!="POD"
  }[5m])
)

CPU usage per pod, breakdown per container

sum by (namespace, pod, container) (
  rate(container_cpu_usage_seconds_total{
    namespace!~"kube-.*|openshift-.*",
    container!="", container!="POD"
  }[5m])
)

CPU usage vs limit (rischio throttling)

sum by (namespace, pod) (
  rate(container_cpu_usage_seconds_total{namespace!~"kube-.*|openshift-.*"}[5m])
)
/
sum by (namespace, pod) (
  container_spec_cpu_quota{namespace!~"kube-.*|openshift-.*"}
  / container_spec_cpu_period{namespace!~"kube-.*|openshift-.*"}
)

CPU throttling (evento certo, non stimato)

rate(container_cpu_cfs_throttled_seconds_total{
  namespace!~"kube-.*|openshift-.*"
}[5m])

% periodi throttled sul totale periodi CFS

sum by (namespace, pod, container) (
  increase(container_cpu_cfs_throttled_periods_total{container!=""}[5m])
)
/
sum by (namespace, pod, container) (
  increase(container_cpu_cfs_periods_total{container!=""}[5m])
)

Memoria

Memory usage per pod (working set, somma container)

sum by (namespace, pod) (
  container_memory_working_set_bytes{
    namespace!~"kube-.*|openshift-.*",
    container!="", container!="POD"
  }
)

Memory usage per container

container_memory_working_set_bytes{
  namespace!~"kube-.*|openshift-.*",
  container!="", container!="POD"
}

Memory RSS (uso reale del processo, non cache)

container_memory_rss{
  namespace!~"kube-.*|openshift-.*",
  container!="", container!="POD"
}

Memory usage vs limit (rischio OOM)

container_memory_working_set_bytes{namespace!~"kube-.*|openshift-.*"}
/
container_spec_memory_limit_bytes{namespace!~"kube-.*|openshift-.*"}

Memory limit non impostato (limit=0, rischio noisy neighbour)

container_spec_memory_limit_bytes{
  namespace!~"kube-.*|openshift-.*", container!=""
} == 0

Confronto requests/limits vs consumo reale

# CPU: uso reale / requests dichiarate
sum by (namespace, pod, container) (
  rate(container_cpu_usage_seconds_total{container!="", container!="POD"}[5m])
)
/ on(namespace, pod, container) group_left
sum by (namespace, pod, container) (
  kube_pod_container_resource_requests{resource="cpu"}
)

# Memoria: uso reale / requests dichiarate
sum by (namespace, pod, container) (
  container_memory_working_set_bytes{container!="", container!="POD"}
)
/ on(namespace, pod, container) group_left
sum by (namespace, pod, container) (
  kube_pod_container_resource_requests{resource="memory"}
)

Utile per capire pod oversized (requests molto più alte del consumo reale → spreco quota nodo) o sottodimensionati (rischio throttling/OOM).