Kubernetes knowledge

 

Master + Cluster (Node 1 + Node 2….)

Master:

1. Kube-apiserver

2. Etcd - key value store

3. Scheduler - distribute work

4. Controller 

a. Node controller - start new container and stop broken one

b. Replication controller

Cluster

1. Kubelet - agent check container status

2. Kube-proxy: communicate between nodes

Etcd 

key - value store

Default port: 2379

Download: https://github.com/etcd-io/etcd/releases/tag/v3.3.13

Extract #tar xzvf <file>

Start #./etcd

Set #./etcdctl set key1 value1

Retrieve #./etcdctl get key1

Etcd in K8s

Save all related information for K8S cluster

(Nodes / Pods / Configs / Secrets / Accounts / Roles / Bindings / Others)

Check #kubectl get pods -n kube-system

Kube - apiserver

Only kube-apiserver to update ETCD 

Create pod process

1. Authenticate User

2. Validate Request

3. Retrieve data

4. Update ETCD

5. Scheduler

6. Node - Kubelet

Check #kubectl get pods --namespace kube-system

Controller Manager

Watch status

Remediate Situation

Node controller --> kube -apiserver --> kubelet (Monitor node)

Node Monitor Period = 5s

Node Monitor Grace Period = 40s (wait 40s to mark node as unreachable)

POD Eviction Timeout = 5m (wait 5m to remove the pods in the node)

Replication Controller / Replica Set --> kube-apiserver --> kubelet (Monitor pod / replicaset)

Using label / selector to monitor the replicaset status

Scheduler 

Decide which pod should deploy to which node

1. Filter Nodes (hardware / software)

2. Rand Nodes

Set pod directly to specific node

In pod.yml add below in spec

nodeName: <specific node name>

https://github.com/tomshenhao/kubernetes-learning/blob/master/pod_define_node.yml

Hard requirement for pods by default

CPU: 0.5

Memory: 256MB

manually define initial hard requirement

spec:

    resources:

        requests:

             memory: "1Gi"

             cpu: 1

cpu min set: 0.1

memory setting

k8s resource limits by default

CPU: 1

memory: 512Mi

manually define max hard requirement

spec:

    resources:

        limits:

              memory: "2Gi"

              cpu: 2

pod use over memory limit --> terminated by K8S 

Monitor

Metrics server - in memory solution

Setup

#git clone https://github.com/kodekloudhub/kubernetes-metrics-server.git

#cd kubernetes-metrics-server

#kubectl create -f .

Check

#kubectl top node

#kubectl top pod

Logs

Check events #kubectl get events

Check log #kubectl logs -f <pod name> <container name>

Label and selector

Define label in yaml file

Search #kubectl get pods --selector <label define>

search with more condition #kubectl get pods --selector <label define>,<label define>

For example

Yaml file

labels:

     app: App1

Selector

#kubectl get pods --selector app=App1

Daemon sets 

Similar as replicaset, but pod will be create to node automatically and remove when node removed.

Use case: monitor solution / log viewer / kube-proxy

Check #kubectl get daemonsets

Create #kubectl create -f <yaml file>

https://github.com/tomshenhao/kubernetes-learning/blob/master/daemonset.yml

Kubelet

Register node

Create PODS

Monitor nodes / pods

Kubeadm does not deploy kubelet

Kube-proxy

Run on each node 

Monitor service, create rule in each node make pod communication.

Kubectl

1. Use --dry-run to test 

2. -o yaml: to output the resource definition in yaml

Yaml File 

Get the yaml file from existing item

kubectl get <item type> <item name> -o yaml --export

Example: kubectl get deploy webapp -o yaml --export

Yaml Lint to verify yaml

www.yamllint.com

K8S must contains

1. apiVersion: string

2. kind: string

3. metadata: dictionary

4. spec: dictionary

Dictionary - unordered

List  - ordered

Using kubectl run command to generate the example yaml file

#kubectl run nginx --image=nginx -o yaml

#kubectl run nginx --image=nginx --replicas=3 -o yaml

Create yaml file from existing pod

#kubectl get pod <pod name> -o yaml > <yaml file>

Replication Controller  (Replica Set)

1. high availability

2. Load balancing & Scaling

https://kubernetes.io/docs/concepts/workloads/controllers/replicaset/

Replication Controller: old technology

Replica Set: new technology 

Create #kubectl create -f <yaml file>

Check #kubectl get replicaset

Check #kubectl describe replicaset

Scale #kubectl scale --replicas=6 -f <yaml file>

Delete #kubectl delete replicaset <replicaset name>

Edit replicaset  directly #kubectl edit replicaset <replicaset name>

If the replicaset broken, but no yaml file.

1. Edit replicaset directly

2. Manually delete the pod

When delete one, it automatically create one

If manually create one, it block the activity. 

Example yaml file

https://github.com/tomshenhao/kubernetes-learning/blob/master/replica.yml

Commands & Arguments

docker file: ubuntu-sleeper

FROM Ubuntu

ENTRYPOINT ["sleep"]

CMD ["5"]

Pod yaml file

spec:

    containers:

    - name: ubuntu-sleeper

      image: ubuntu-sleeper

      command:

      - "sleep"

      - "5000"

Monitor tool

https://github.com/kodekloudhub/kubernetes-metrics-server.git