knowledge/categories/devops/kubernetes.md

Kubernetes

Terminology

• Pod
  • Single container
  • Multiple containers
    • share ip
    • share volumes
• Node
• Service
  • ClusterIP (default)
  • NodePort （ access by port）
  • LoadBalancer (external ip)
  • ExternalName dns
• Deployment
• Replica Set
• Service rolling update
• DaemonSet
  • running a cluster storage daemon, such as glusterd, ceph, on each node
  • running a logs collection daemon on every node, such as fluentd or logstash.
  • running a node monitoring daemon on every node, such as Prometheus Node Exporter, collectd, New Relic agent, or Ganglia gmond.
• StatefulSets
  • Ordered, graceful deployment and scaling
  • Stable, persistent storage.
  • Stable, unique network identifiers.
  • Ordered, graceful deletion and termination.
• Proxy
• DNS
  • service.namespace
• Secrets
• Persistent volumes
  • Available
    • a free resource that is not yet bound to a claim
  • Bound
    • the volume is bound to a claim
  • Released
    • the claim has been deleted, but the resource is not yet reclaimed by the cluster
  • Failed
    • the volume has failed its automatic reclamation

Several ways to access a service inside Kubernetes cluster

• ClusterIP, Can only be accessed inside the cluster. You can access inside kubectl proxy
• NodePort, You can access the service by a specific public port between 30000–32767, if the cluster server IP changes, then your service' endpoint url changes.
• LoadBalancer, it's kind of external service, all the requests go through the loadbalancer. Every service needs a new IP for this.
• Ingress, can be shared among multiple services, you can have different types of ingress controllers: gec, nginx, contour,istio and so on.

Links:

• Kubernetes NodePort vs LoadBalancer vs Ingress? When should I use what?

Zero down time deployment

Set strategy type to RollingUpdate instead of Recreate

spec:
  replicas: 1
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 50%
      maxSurge: 1

Some useful commands

kubectl config set-context context-name --namespace default-namespace-name #default active context
kubectl config view #view config
kubectl get nodes #get nodes
kubectl get namespaces #get all namespaces
kubectl get pods #get pods
kubectl get deployments #get deployments
kubectl logs -f <pod> --tail 200 # tail logs from some pod
kubectl get services #get services
kubectl get ingress #get ingresses
kubectl get hpa # get horizontal auto scaling policies
kubectl get all # get all kinds of units
kubectl get secrets <secret-id> -o yaml #view a secret details with yaml format, fields are encrypted with base64

Kubernetes Dashboard

Deploy Dashboard

kubectl create -f https://raw.githubusercontent.com/kubernetes/dashboard/master/src/deploy/recommended/kubernetes-dashboard.yaml

Access dashboard

# First fetch cluster info, make sure cluster is running properly
kubectl cluster-info

# Normally the url for dashboard would be
http://your_ip:8080/api/v1/namespaces/kube-system/services/https:kubernetes-dashboard:/proxy/

Configure Kubectl to access remote Kubernetes cluster

Here is a very simple config which use HTTP and not secure, should only for local testing purpose

Config

Having a yaml file called config-demo in your current folder

apiVersion: v1
clusters:
- cluster:
    server: http://your_cluster_ip:8080
  name: development
contexts:
- context:
    cluster: development
    namespace: dev
    user: developer
  name: dev
current-context: dev
kind: Config
preferences: {}
users:
- name: developer

Connect to remote cluster

kubectl get all --kubeconfig=config-demo --all-namespaces

For long term usage, you will need to copy the content to your ~/.kube/config

TODO: Setup SSL connection

Helm

Add Kubernetes yaml template engine and the package manager for Kubernetes

How to use

helm init
helm upgrade --install -f abc/values-staging.yaml some-name ./abc
# abc/values-staging.yaml the value file
# some-name the release name
# abc the template folder
helm delete --purge mqtt  # mqtt the release name

Install 3rd party packages

helm repo add gitlab https://charts.gitlab.io/ # add remote repo
helm repo update # update index
helm install mirantisworkloads/vernemq

Tips

• Normally when you just updated the configmap the deployment or statefulset pod wouldn't updated, but you can add a label to deployment/statefulset yaml, when the value changes the pods will be recreated

template:
  metadata:
    labels:
      app: vernemq
      configmapVersion: "{{ .Release.Revision }}"

Deploy tools comparsion

https://blog.hasura.io/draft-vs-gitkube-vs-helm-vs-ksonnet-vs-metaparticle-vs-skaffold-f5aa9561f948/

Horizontal scaling

V1

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: worker-auto-scaling
  namespace: x-prod
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: <deployment-name>
  minReplicas: 1
  maxReplicas: 2
  targetCPUUtilizationPercentage: 75 # trigger point

V2

apiVersion: autoscaling/v2beta2
kind: HorizontalPodAutoscaler
metadata:
  name: php-apache
  namespace: default
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: php-apache
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50
  - type: Pods
    pods:
      metric:
        name: packets-per-second
      target:
        type: AverageValue
        averageValue: 1k
  - type: Object
    object:
      metric:
        name: requests-per-second
      describedObject:
        apiVersion: networking.k8s.io/v1beta1
        kind: Ingress
        name: main-route
      target:
        type: Value
        value: 10k

Limit CPU and memory for pods

ports:
- name: db-port
  containerPort: 2345
  protocol: TCP
resources:
requests:
  cpu: 30m
  memory: 64Mi
limits:
  cpu: 200m
  memory: 256Mi

Tools

Kubectx and kubens

Switch faster between clusters and namespaces in kubectl https

• kubectx for switching contexts
• kubens for switching namespaces

Github page: https://github.com/ahmetb/kubectx