Host metrics

Host metrics

Can containers use all resources? Yes, but they should not!

Some reservations are necessary

Capacity must be divided between system, cluster and containers

Operating system

Reserved for system services

Kubernetes

Reserved for cluster components

Allocatable resources

Allocatable = Capacity - System - Kubernetes

Reservations in Managed Kubernetes

Overview of AWS, Azure and Google Cloud

Larger VMs have less overhead

More VMs provide more availability

Further reading

Instance calculator for cloud providers

Read reservations from managed cluster

Configure Reservations

kubelet flags:

--system-reserved=memory=500Mi,cpu=100m
--kube-reserved=memory=500Mi,cpu=100m

Implementation depends on deployment

For example in kind:

kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
nodes:
- role: control-plane
  kubeadmConfigPatches:
  - |
    kind: InitConfiguration
    nodeRegistration:
      kubeletExtraArgs:
        system-reserved: memory=12Gi,cpu=2000m
        kube-reserved: memory=100Mi,cpu=100m

Host metrics collection

node-exporter collects host metrics…

…and exports them for scraping

Metrics include CPU, memory, disk, network and a lot more!

Some are disabled but the defaults are reasonable

Demo

Start Kubernetes API proxy and read metrics endpoint:

kubectl proxy

H=localhost:8001
NS=kube-system
SVC=node-exporter-prometheus-node-exporter
curl -s $H/api/v1/namespaces/$NS/services/$SVC:metrics/proxy/metrics \
| grep node_cpu_seconds_total