[cloud native | kubernetes] in depth understanding of deployment (VIII)

​ Deepen understanding Deployment

​

One 、 What is? Deployment

• One Deployment by Pods and ReplicaSets Provide declarative update capabilities .

• You are responsible for describing Deployment Medium _ Target state _, and Deployment controller （Controller） Change at a controlled rate The actual state , Turn it into Expected state ; Control cycle . for(){ xxx controller.spec()}

• Don't manage Deployment Owned ReplicaSet

• When we deploy an application, we usually don't write directly Pod, Instead, deploy a Deployment

• Deploy Preparation of the statute Deployments | Kubernetes

Two 、Deployment establish

• The basic format

  • .metadata.name Appoint deploy name

  • replicas Specify the number of copies

  • selector Specify the matching Pod Templates .

  • template Make a statement Pod Templates

Write a Deployment Of yaml give Pod Self healing and failover capabilities

• Check the... In the cluster Deployment when , The fields displayed are ：

  • NAME Lists the Deployment The name of .

  • READY Shows the available copy Count . The mode of display is “ The number of ready / Expected number ”.

  • UP-TO-DATE Shows the number of copies that have been updated to achieve the desired state .

  • AVAILABLE Shows the number of copies of the app available to users .

  • AGE Shows when the application is running .

• ReplicaSet The output contains the following fields ：

  • NAME List... In the namespace ReplicaSet The name of ;

  • DESIRED Shows the expected number of copies of the app , Creating Deployment The value defined when . This is the expected state ;

  • CURRENT Displays the number of copies in the current running state ;

  • READY Shows how many copies of the app can serve users ;

  • AGE Shows how long the app has been running .

  • Be careful ：ReplicaSet The name of is always formatted as [Deployment name ]-[ Random string ]. Where the random string is used pod-template-hash Randomly generated as seeds .

One Deploy Produce three

• Deployment resources

• replicaset resources

• Pod resources

Deployment control RS,RS control Pod Number of copies ReplicaSet： It only provides the control function of the number of copies Deployment： Every time a new version is deployed, a new replica set is created , Use him to record status , Rollback is also specified directly rs take effect

3、 ... and 、Deployment Renewal mechanism

• Only when the Deployment Pod Templates （ namely .spec.template） When there is a change , for example The label or container image of the template is updated , Will trigger Deployment go online . Other updates （ If yes Deployment Perform the expansion and reduction operation ） It doesn't trigger an online action .

• Online action principle ： Create a new rs, When ready , Replace old rs（ Will not be deleted at this time , because **revisionHistoryLimit** Specifies how many versions to keep ）

• frequently-used kubectl command

################ to update #################################
#kubectl  set image  deployment Resource name    Container name = Mirror name 
kubectl set image deployment.apps/nginx-deployment php-redis=tomcat:8 --record
## yaml Extract all the key fields that can be updated hash.
web---- /hello
postman   aservice- /hello

# Or modify the definition directly 
kubectl edit deployment.v1.apps/nginx-deployment
# Check the status 
kubectl rollout status deployment.v1.apps/nginx-deployment

################ View the history and roll back ####################################
# Check out the update history - See if the total number of history records we set is effective 
kubectl rollout history deployment.v1.apps/nginx-deployment
# Roll back 
kubectl rollout undo deployment.v1.apps/nginx-deployment --to-revision=2

############### Cumulative update ##############
# Pause recording version 
kubectl rollout pause deployment.v1.apps/nginx-deployment
# Multiple update operations .
## For example, the resource limit has been updated 
kubectl set resources deployment.v1.apps/nginx-deployment -c=nginx --limits=cpu=200m,memory=512Mi
## For example, the image version has been updated 
kubectl set image deployment.apps/nginx-deployment php-redis=tomcat:8
## Continue operation several times 
## See if the historical version records any changes 
kubectl rollout history deployment.v1.apps/nginx-deployment
# Make multiple cumulative effects 
kubectl rollout resume deployment.v1.apps/nginx-deployment

​

1、 Scaling （Proportional Scaling）

maxSurge（ Maximum increment ）： How many instances should be added besides the current number .

maxUnavailable（ Maximum unusable amount ）： Number of unavailable instances during rolling update .

2、HPA（ Dynamic expansion and contraction ）

Concept ：Pod Horizontal automatic expansion and contraction | Kubernetes

actual combat ：HorizontalPodAutoscaler rehearse | Kubernetes

​

2.1、 You need to install metrics-server

GitHub - kubernetes-sigs/metrics-server: Scalable and efficient source of container resource metrics for Kubernetes built-in autoscaling pipelines.

2.2、 Installation steps

apiVersion: v1
kind: ServiceAccount
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server
    rbac.authorization.k8s.io/aggregate-to-admin: "true"
    rbac.authorization.k8s.io/aggregate-to-edit: "true"
    rbac.authorization.k8s.io/aggregate-to-view: "true"
  name: system:aggregated-metrics-reader
rules:
- apiGroups:
  - metrics.k8s.io
  resources:
  - pods
  - nodes
  verbs:
  - get
  - list
  - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server
  name: system:metrics-server
rules:
- apiGroups:
  - ""
  resources:
  - pods
  - nodes
  - nodes/stats
  - namespaces
  - configmaps
  verbs:
  - get
  - list
  - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server-auth-reader
  namespace: kube-system
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: extension-apiserver-authentication-reader
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server:system:auth-delegator
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:auth-delegator
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: system:metrics-server
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:metrics-server
subjects:
- kind: ServiceAccount
  name: metrics-server
  namespace: kube-system
---
apiVersion: v1
kind: Service
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
spec:
  ports:
  - name: https
    port: 443
    protocol: TCP
    targetPort: https
  selector:
    k8s-app: metrics-server
---
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
spec:
  selector:
    matchLabels:
      k8s-app: metrics-server
  strategy:
    rollingUpdate:
      maxUnavailable: 0
  template:
    metadata:
      labels:
        k8s-app: metrics-server
    spec:
      containers:
      - args:
        - --cert-dir=/tmp
        - --kubelet-insecure-tls
        - --secure-port=4443
        - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname
        - --kubelet-use-node-status-port
        image: registry.cn-hangzhou.aliyuncs.com/lanson_k8s_images/metrics-server:v0.4.3
        imagePullPolicy: IfNotPresent
        livenessProbe:
          failureThreshold: 3
          httpGet:
            path: /livez
            port: https
            scheme: HTTPS
          periodSeconds: 10
        name: metrics-server
        ports:
        - containerPort: 4443
          name: https
          protocol: TCP
        readinessProbe:
          failureThreshold: 3
          httpGet:
            path: /readyz
            port: https
            scheme: HTTPS
          periodSeconds: 10
        securityContext:
          readOnlyRootFilesystem: true
          runAsNonRoot: true
          runAsUser: 1000
        volumeMounts:
        - mountPath: /tmp
          name: tmp-dir
      nodeSelector:
        kubernetes.io/os: linux
      priorityClassName: system-cluster-critical
      serviceAccountName: metrics-server
      volumes:
      - emptyDir: {}
        name: tmp-dir
---
apiVersion: apiregistration.k8s.io/v1
kind: APIService
metadata:
  labels:
    k8s-app: metrics-server
  name: v1beta1.metrics.k8s.io
spec:
  group: metrics.k8s.io
  groupPriorityMinimum: 100
  insecureSkipTLSVerify: true
  service:
    name: metrics-server
    namespace: kube-system
  version: v1beta1
  versionPriority: 100

​

• kubectl apply that will do

• All runnning use

  • kubectl top nodes --use-protocol-buffers

  • kubectl top pods --use-protocol-buffers

2.3、 To configure hpa test

###  The test image  registry.cn-hangzhou.aliyuncs.com/lanson_k8s_images/php-hpa:latest

## Applied yaml Have done well 
apiVersion: v1
kind: Service
metadata:
  name: php-apache
spec:
  ports:
  - port: 80
    protocol: TCP
    targetPort: 80
  selector:
    run: php-apache
---
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    run: php-apache
  name: php-apache
spec:
  replicas: 1
  selector:
    matchLabels:
      run: php-apache
  template:
    metadata:
      creationTimestamp: null
      labels:
        run: php-apache
    spec:
      containers:
      - image: registry.cn-hangzhou.aliyuncs.com/lanson_k8s_images/php-hpa:latest
        name: php-apache
        ports:
        - containerPort: 80
        resources:
          requests:
            cpu: 200m

##hpa To configure  hpa.yaml
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: php-apache
spec:
  maxReplicas: 10
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: php-apache
  targetCPUUtilizationPercentage: 50
  
#3、 Stress test 
kubectl run -i --tty load-generator --image=busybox /bin/sh

# Enter and hit the following command 
 kubectl run -i --tty load-generator --rm --image=busybox --restart=Never -- /bin/sh -c "while sleep 0.01; do wget -q -O- http://php-apache; done"

​

3、Canary（ Canary deployment ）

3.1、 Blue and green deployment VS Canary deployment

Blue and green deployment

​

Canary deployment

​

3.2、 A simple test of Canary

Use this image to test registry.cn-hangzhou.aliyuncs.com/lanson_k8s_images/nginx-test This mirror image docker run When -e msg=aaaa, Visit this nginx Page is to see aaaa

Step principle

• Prepare one Service, Load balancing Pod

• Prepare version v1 Of deploy, Prepare version v2 Of deploy

4、Deployment Status and troubleshooting

Deployments | Kubernetes

​

​