TGIK8s 017 notes

general notes and links

Things to cover:
* Installing heapster and what it does
* requests vs. limits, compressible vs. incompressible
* Running kuard without any limits, show memory in heapster
* Resource QoS: Guaranteed, Burstable, BestEffort
* [coming soon] Pod priority/preemption
* Quota
* LimitRange admission controller

Links:
* Main docs page on resources: https://kubernetes .io/docs/concepts/configuration/manage-compute-resources-container/
* LimitRange - https://kubernetes.io/docs/tasks/administer-cluster/memory-default-namespace/
* Quota - https://kubernetes.io/docs/tasks/administer-cluster/quota-memory-cpu-namespace/

* Good blog post on QoS: https://medium.com/google-cloud/quality-of-service-class-qos-in-kubernetes-bb76a89eb2c6
* Pod priority and preemption: https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/
* Kubelet out of resource handling: https://kubernetes.io/docs/tasks/administer-cluster/out-of-resource/#node-oom-behavior

---
# Install Heapster
# https://github.com/kubernetes/heapster/blob/master/docs/influxdb.md
git clone https://github.com/kubernetes/heapster.git
cd heapster
kubectl apply -f deploy/kube-config/influxdb/
kubectl apply -f deploy/kube-config/rbac/heapster-rbac.yaml
# http://localhost:8001/api/v1/proxy/namespaces/kube-system/services/monitoring-grafana:80/
# If using dashboard, kill that pod to have it pick up that heapster is available right away.

kuard-g.yaml

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  labels:
    run: kuard-g
  name: kuard-g
spec:
  replicas: 6
  selector:
    matchLabels:
      run: kuard-g
  template:
    metadata:
      labels:
        run: kuard-g
    spec:
      containers:
      - image: gcr.io/kuar-demo/kuard-amd64:2
        name: kuard-g
        imagePullPolicy: Always
        resources:
            limits:
              memory: "2Gi"
              cpu: "200m"
      nodeSelector:
        type: test

re-port-forward.sh

#! /bin/bash

set +m

while :
do
    kubectl port-forward $(kubectl get pod -l run=$1 -o jsonpath='{.items[0].metadata.name}') $2:8080 >&/dev/null &
    TASK_PID=$!
    sleep 10s
    { kill $TASK_PID && wait; }>& /dev/null
done