Kubernetes¶

Resources¶

Tools¶

K9s¶

Kubectl TUI

Install

brew install k9s

To go in

k9s

Commands

select namespace
- :ns
see other thing
- :deployment/service/pods/etc.
quit
- :q

CD¶

ArgoCD

install kubectl¶

doc

install kubectl binary

curl -LO "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/amd64/kubectl"

install kubectl

sudo install -o root -g root -m 0755 kubectl /usr/local/bin/kubectl

test

kubectl version --client

Apply config file¶

Write config files in yaml.

operations

create
apply
replace
patch

To create

kubectl create -f <config_file>.yaml

To create/update

kubectl apply -f <config_file>.yaml

create vs. apply

labels¶

Labels are key-value pair. You can assign labels via command line or define in yaml. See the doc.

roles¶

Role is just another label with key = kubernetes.io/role and value = <your_role>.

commands¶

get component status
- kubectl get cs (deprecated)
show info about a pod
- kubectl get pod <pod_name>
show info about a node
- kubectl get node <node_name>

show nodes¶

# show ip
kubectl get nodes -o wide
# show labels
kubectl get nodes --show-labels
# show nodes with certain label
kubectl get nodes -l <key>=<value>

assign pod to node¶

https://kubernetes.io/docs/tasks/configure-pod-container/assign-pods-nodes/

label your node -> specify nodeSelector in your yaml

Note that in Deployment, it should be in template.spec

Pulling image¶

pull image from private docker registry¶

doc

login to docker if haven't

docker login

see your config.json

cat ~/.docker/config.json

create kubectl create

kubectl create secret generic <secret_name> \
    --from-file=.dockerconfigjson=<path/to/.docker/config.json> \
    --type=kubernetes.io/dockerconfigjson

supply secret in config file

apiVersion: v1
kind: Pod
metadata:
  name: private-reg
spec:
  containers:
  - name: private-reg-container
    image: <your-private-image>
  imagePullSecrets:
  - name: <secret_name>

create pod¶

kubectl create -f <k8s_config_file>

force recreate pods

kubectl replace --force -f <k8s_config_file>

delete pod¶

kubectl delete -f <k8s_config_file>

kubectl delete pods <pod_name>

limit resources¶

cpu & memory

port forwarding¶

kubectl port-forward <pod_name> <port>:<port>

pod log¶

kubectl logs <pod_name>

show pods¶

kubectl get pod

more info (like ip & node run on)

kubectl get pods -o wide

Go into a pod¶

kubectl exec --stdin --tty <pod_name> -- /bin/bash

Namespace¶

For segratating your resources (e.g. different apps)

Get things from all namespaces¶

If you don't supply a namespace, the get commands will only show the resources in your current namespace. To get the resources under all namespaces, use the --all-namespaces flag.

e.g.

kubectl get pods --all-namespaces

View all namespace¶

kubectl get namespace

ns for short

View currently used namespace¶

kubectl config view | grep namespace

Set namespace¶

kubectl config set-context --current --namespace=<your_namespace>

-n for short

To unset

kubectl config set-context --current --namespace=""

See your cronjob

kubectl get cronjob

deployment¶

spec.template is the definition of the pod

update deployment¶

update image

kubectl set image deployment/<deployment_name> <pod_name>=<new_image>

modify directly

kubectl edit deployment/<deployment_name>

service¶

handle port forwarding

targetPort
- the port inside the pod
nodePort
- the port of the kubernetes node

go to http://<node_ip>:<nodePort> to see your app

internal service url¶

<service_name>.<namespace>.svc.cluster.local:<service_port>

Ingress¶

Doc

Use ingress to map host/path to your service. Note that the host you define must be pointing to an ip pointing to your cluster, e.g. your load balancer or ingress controller

kind¶

It's an alternative to minikube. Kind uses Docker container while minikube uses VM. See the official page and the comparison.

Install¶

Install Docker if you haven't.

Mac¶

brew install kind

Install with binary¶

install binary

curl -Lo ./kind https://kind.sigs.k8s.io/dl/v0.11.1/kind-linux-amd64
chmod +x ./kind
sudo mv ./kind /usr/local/bin/kind

check

kind version

create a cluster¶

kind create cluster

check if it's running

docker ps

kind get clusters1

kubectl get nodes

kind commands¶

create cluster
- kind create cluster
- params
  - --name
    - name of the cluster
    - default name = kind
  - --config
    - create with config yaml
delete cluster
- kind delete cluster
  - --name
    - name of the cluster
    - default name = kind
list clusters
- kind get clusters
list nodes
- kind get nodes
cluster detals
- kubectl cluster-info --context kind-kind

config file¶

doc

Use InitConfiguration to label 1st control plane node, JoinConfiguration for others. See Kubeadm Config Patches (related Github issue).

kind: Cluster
apiVersion: kind.x-k8s.io/v1alpha4
name: test
nodes:
- role: control-plane
  kubeadmConfigPatches:
  - |
    kind: InitConfiguration
    nodeRegistration:
      kubeletExtraArgs:
        node-labels: "name=edge1"
- role: worker
  kubeadmConfigPatches:
  - |
    kind: JoinConfiguration
    nodeRegistration:
      kubeletExtraArgs:
        node-labels: "name=edge1"

create a local container registry¶

create a local docker registry and a kind cluster

https://kind.sigs.k8s.io/docs/user/local-registry/

troubleshooting¶

error when creating a cluster with many nodes¶

If you can successfully create a cluster with less nodes, then the problem might be Pod errors due to “too many open files”.

To solve it, go to /etc/sysctl.conf and add/update the following lines:

fs.inotify.max_user_watches = 524288
fs.inotify.max_user_instances = 512

In my case, my fs.inotify.max_user_watches is already 1048576, so I only add the second line.

related github issue

using webcam¶

      containers:
        - name: <name>
          image: <image>
          volumeMounts:
          - mountPath: /dev/video0
            name: dev-video0
          securityContext:
            privileged: true
      volumes:
        - name: dev-video0
          hostPath:
            path: /dev/video0

https://stackoverflow.com/a/59291859/15493213

Troubleshooting¶

use describe on the failing node/pod to see more info first

1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate¶

Scenario: You make a pod run on a control plane / master node but it keeps pending. When you describe it, 1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate is shown.

Solution: Remove the taint on the node

kubectl taint nodes --all node-role.kubernetes.io/master-
# or
kubectl taint nodes <node_name> node-role.kubernetes.io/master-

the - is for removing

ref:

The connection to the server localhost:8080 was refused - did you specify the right host or port?¶

Happens when you run any kubectl <xxx> command.

Easy fix: delete your cluster and start again

If you're in local and you use kind, kind get clusters and kind delete cluster --name <cluster_name>.

no matches for kind "CronJob" in version "batch/v1beta1"¶

Your k8s version >= 1.21, so use this instead

apiVersion: batch/v1

See https://cloud.google.com/kubernetes-engine/docs/how-to/cronjobs