20210906 (월) deployment strategies, statefulset

 

  Deployment

    Replicaset을 포함하는 상위 컨트롤러로 일정 갯수의 복제본을 유지시켜주는 컨트롤러

  배포 전략 (Deployment Strategy)

    RollingUpdate

    ReCreate

 

### deployments 생성때부터 --record 옵션으로 revision 기록 하기 ###

vagrant@kube-control1:~/work/20210906$ cat myapp-deploy-v1.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: myapp-deploy
  labels:
    app: myapp-deploy
spec:
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 1
      maxSurge: 2
  minReadySeconds: 15
  replicas: 3
  selector:
    matchLabels:
      app: myapp-deploy
  template:
    metadata:
      labels:
        app: myapp-deploy
    spec:
      containers:
      - name: myapp
        image: devops2341/go-myweb:v1
        ports:
        - containerPort: 8080
          protocol: TCP

vagrant@kube-control1:~/work/20210906$ kubectl create -f myapp-deploy-v1.yaml --record
deployment.apps/myapp-deploy created
vagrant@kube-control1:~/work/20210906$ kubectl rollout history deployment
deployment.apps/myapp-deploy
REVISION  CHANGE-CAUSE
1         kubectl create --filename=myapp-deploy-v1.yaml --record=true

### 이미지 교체
vagrant@kube-control1:~/work/20210906$ kubectl set image deployment myapp-deploy myapp=devops2341/go-myweb:v2 --record
deployment.apps/myapp-deploy image updated

### 이미지 교체 이력 확인 (--record 옵션)
vagrant@kube-control1:~/work/20210906$ kubectl rollout history deployments
deployment.apps/myapp-deploy
REVISION  CHANGE-CAUSE
1         kubectl create --filename=myapp-deploy-v1.yaml --record=true
2         kubectl set image deployment myapp-deploy myapp=devops2341/go-myweb:v2 --record=true

### 오타 등으로 이미지교체에 실패했을때, pod는 maxUnavailable 항목에 지정한
#### 1개 pod 제외한 나머지 pod는 정상적인 running상태를 유지한다.

vagrant@kube-control1:~/work/20210906$ kubectl get pods
NAME                            READY   STATUS             RESTARTS   AGE
myapp-deploy-65bdb76f48-9k4vs   1/1     Running            0          3m6s
myapp-deploy-65bdb76f48-vhzb6   1/1     Running            0          3m6s
myapp-deploy-7b88d989c8-2jlk5   0/1     ImagePullBackOff   0          44s
myapp-deploy-7b88d989c8-bw2mp   0/1     ImagePullBackOff   0          44s
myapp-deploy-7b88d989c8-w6bfn   0/1     ImagePullBackOff   0          44s

vagrant@kube-control1:~/work/20210906$ kubectl rollout history  deployments
deployment.apps/myapp-deploy
REVISION  CHANGE-CAUSE
1         kubectl create --filename=myapp-deploy-v1.yaml --record=true
2         kubectl set image deployment myapp-deploy myapp=devops2341/go-myweb:v2 --record=true
3         kubectl set image deployment myapp-deploy myapp=devops2341/v3 --record=true

vagrant@kube-control1:~/work/20210906$ kubectl rollout undo deployment myapp-deploy --to-revision 2
deployment.apps/myapp-deploy rolled back
vagrant@kube-control1:~/work/20210906$ kubectl get pods
NAME                            READY   STATUS    RESTARTS   AGE
myapp-deploy-65bdb76f48-47fkt   1/1     Running   0          9s
myapp-deploy-65bdb76f48-9k4vs   1/1     Running   0          7m9s
myapp-deploy-65bdb76f48-vhzb6   1/1     Running   0          7m9s

 

---

 

  Statefulset

    pod의 복제본 관리

    pod의 고유성 (이름, 네트워크, 스토리지)

    pod의 순서 (Order)

 

  Statefulset 주의사항

    pod에 사용할 스토리지는 PVC를 통해서만 할당 가능함

    (→ 미리 PV를 생성하거나 볼륨을 동적 프로비저닝이 가능해야함.)

    pod는 고유성을 가지므로 각각의 pod에 접근 시 Headless 서비스를 이용

    pod의 이름은 Statefulset_NAME-# 형식 (# 0부터 순서대로 숫자부여)

    pod의 Kubernetes 도메인 주소는 POD_NAME.DOMAIN_NAME.svc

 

 

vagrant@kube-control1:~/work/20210906$ cat myapp-sts.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: myapp-sts
spec:
  selector:
    matchLabels:
      app: myapp-sts
  serviceName: myapp-svc-headless
  replicas: 2
  template:
    metadata:
      labels:
        app: myapp-sts
    spec:
      containers:
      - name: myapp
        image: devops2341/go-myweb:latest
        ports:
        - containerPort: 8080
          protocol: TCP
vagrant@kube-control1:~/work/20210906$ cat myapp-svc-headless.yaml
apiVersion: v1
kind: Service
metadata:
  name: myapp-svc-headless
  labels:
    app: myapp-svc-headless
spec:
  clusterIP: None
  ports:
  - name: http
    port: 8080
    targetPort: 8080
  selector:
    app: myapp-sts

 

vagrant@kube-control1:~/work/20210906$ kubectl get statefulsets NAME        READY   AGE myapp-sts   2/2     48s vagrant@kube-control1:~/work/20210906$ kubectl get pods NAME                            READY   STATUS    RESTARTS   AGE myapp-deploy-678f76c6d7-dk4sl   1/1     Running   0          136m myapp-deploy-678f76c6d7-f5qz6   1/1     Running   0          136m myapp-deploy-678f76c6d7-pn9kt   1/1     Running   0          136m myapp-sts-0                     1/1     Running   0          52s myapp-sts-1                     1/1     Running   0          20s vagrant@kube-control1:~/work/20210906$ kubectl describe statefulset myapp-sts Name:               myapp-sts Namespace:          default CreationTimestamp:  Mon, 06 Sep 2021 06:19:19 +0000 Selector:           app=myapp-sts Labels:             <none> Annotations:        <none> Replicas:           2 desired | 2 total Update Strategy:    RollingUpdate   Partition:        0 Pods Status:        2 Running / 0 Waiting / 0 Succeeded / 0 Failed Pod Template:   Labels:  app=myapp-sts   Containers:    myapp:     Image:        devops2341/go-myweb:latest     Port:         8080/TCP     Host Port:    0/TCP     Environment:  <none>     Mounts:       <none>   Volumes:        <none> Volume Claims:    <none> Events:   Type    Reason            Age   From                    Message   ----    ------            ----  ----                    -------   Normal  SuccessfulCreate  95s   statefulset-controller  create Pod myapp-sts-0 in StatefulSet myapp-sts successful   Normal  SuccessfulCreate  63s   statefulset-controller  create Pod myapp-sts-1 in StatefulSet myapp-sts successful vagrant@kube-control1:~/work/20210906$ kubectl get services NAME                 TYPE           CLUSTER-IP      EXTERNAL-IP      PORT(S)        AGE kubernetes           ClusterIP      10.96.0.1       <none>           443/TCP        13d myapp-svc-deploy     LoadBalancer   10.111.36.106   192.168.200.11   80:30834/TCP   152m myapp-svc-headless   ClusterIP      None            <none>           8080/TCP       2m1s vagrant@kube-control1:~/work/20210906$ kubectl get pods -o wide NAME                            READY   STATUS    RESTARTS   AGE     IP                NODE         NOMINATED NODE   READINESS GATES myapp-deploy-678f76c6d7-dk4sl   1/1     Running   0          137m    192.168.119.169   kube-node3   <none>           <none> myapp-deploy-678f76c6d7-f5qz6   1/1     Running   0          137m    192.168.233.217   kube-node2   <none>           <none> myapp-deploy-678f76c6d7-pn9kt   1/1     Running   0          137m    192.168.9.70      kube-node1   <none>           <none> myapp-sts-0                     1/1     Running   0          2m29s   192.168.9.69      kube-node1   <none>           <none> myapp-sts-1                     1/1     Running   0          117s    192.168.119.168   kube-node3   <none>           <none> vagrant@kube-control1:~/work/20210906$ curl http://192.168.9.69 curl: (7) Failed to connect to 192.168.9.69 port 80: Connection refused vagrant@kube-control1:~/work/20210906$ curl http://192.168.9.69:8080 Hello World! myapp-sts-0 vagrant@kube-control1:~/work/20210906$ kubectl scale statefulset myapp-sts --replicas 4 statefulset.apps/myapp-sts scaled vagrant@kube-control1:~/work/20210906$ kubectl get pods NAME                            READY   STATUS    RESTARTS   AGE myapp-deploy-678f76c6d7-dk4sl   1/1     Running   0          139m myapp-deploy-678f76c6d7-f5qz6   1/1     Running   0          139m myapp-deploy-678f76c6d7-pn9kt   1/1     Running   0          139m myapp-sts-0                     1/1     Running   0          4m3s myapp-sts-1                     1/1     Running   0          3m31s myapp-sts-2                     1/1     Running   0          19s myapp-sts-3                     1/1     Running   0          14s vagrant@kube-control1:~/work/20210906$ kubectl scale statefulset myapp-sts --replicas 2 statefulset.apps/myapp-sts scaled vagrant@kube-control1:~/work/20210906$ kubectl get pods NAME                            READY   STATUS    RESTARTS   AGE myapp-deploy-678f76c6d7-dk4sl   1/1     Running   0          140m myapp-deploy-678f76c6d7-f5qz6   1/1     Running   0          140m myapp-deploy-678f76c6d7-pn9kt   1/1     Running   0          140m myapp-sts-0                     1/1     Running   0          4m54s myapp-sts-1                     1/1     Running   0          4m22s

vagrant@kube-control1:~/work/20210906$ kubectl run nettool -it --image devops2341/network-multitool:v1 --rm bash If you don't see a command prompt, try pressing enter. bash-5.1# host myapp-sts-0.myapp-svc-headless myapp-sts-0.myapp-svc-headless.default.svc.cluster.local has address 192.168.9.69 bash-5.1# host myapp-sts-1.myapp-svc-headless myapp-sts-1.myapp-svc-headless.default.svc.cluster.local has address 192.168.119.168 bash-5.1# host myapp-svc-headless myapp-svc-headless.default.svc.cluster.local has address 192.168.119.168 myapp-svc-headless.default.svc.cluster.local has address 192.168.9.69 myapp-svc-headless.default.svc.cluster.local has address 192.168.233.227 bash-5.1# curl http://myapp-sts-0.myapp-svc-headless:8080 Hello World! myapp-sts-0 bash-5.1# curl http://myapp-sts-1.myapp-svc-headless:8080 Hello World! myapp-sts-1

반응형

 

vagrant@kube-control1:~/work/20210906$ cat myapp-svc-headless2.yaml
apiVersion: v1
kind: Service
metadata:
  name: myapp-svc-headless2
  labels:
    app: myapp-svc-headless2
spec:
  clusterIP: None
  ports:
  - name: http
    port: 8080
    targetPort: 8080
  selector:
    app: myapp-sts-vol
vagrant@kube-control1:~/work/20210906$ cat myapp-sts-volume.yaml
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: myapp-sts-vol
spec:
  selector:
    matchLabels:
      app: myapp-sts-vol
  serviceName: myapp-svc-headless2
  replicas: 2
  template:
    metadata:
      labels:
        app: myapp-sts-vol
    spec:
      containers:
      - name: myapp
        image: devops2341/go-myweb:alpine
        ports:
        - containerPort: 8080
          protocol: TCP
        volumeMounts:
        - name: myapp-data
          mountPath: /data
  volumeClaimTemplates:
  - metadata:
      name: myapp-data
    spec:
      accessModes: ["ReadWriteOnce"]
      resources:
        requests:
          storage: 1Gi
      storageClassName: rook-ceph-block

vagrant@kube-control1:~/work/20210906$

 

vagrant@kube-control1:~/work/20210906$ kubectl get pods NAME                            READY   STATUS              RESTARTS   AGE myapp-deploy-678f76c6d7-dk4sl   1/1     Running             0          173m myapp-deploy-678f76c6d7-f5qz6   1/1     Running             0          173m myapp-deploy-678f76c6d7-pn9kt   1/1     Running             0          173m myapp-sts-0                     1/1     Running             0          38m myapp-sts-1                     1/1     Running             0          37m myapp-sts-2                     1/1     Running             0          33m myapp-sts-vol-0                 1/1     Running             0          43s myapp-sts-vol-1                 0/1     ContainerCreating   0          15s nettool                         1/1     Running             0          31m vagrant@kube-control1:~/work/20210906$ kubectl get persistentvolumeclaims NAME                         STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS      AGE myapp-data-myapp-sts-vol-0   Bound    pvc-d0a53c0b-135f-4226-a2e0-89c961bb6a37   1Gi        RWO            rook-ceph-block   57s myapp-data-myapp-sts-vol-1   Bound    pvc-e8f04ef5-26f5-4a95-a15e-0defc3b6fd9d   1Gi        RWO            rook-ceph-block   29s vagrant@kube-control1:~/work/20210906$ kubectl describe pods myapp-sts-vol-0 Name:         myapp-sts-vol-0 Namespace:    default Priority:     0 Node:         kube-node1/192.168.200.21 Start Time:   Mon, 06 Sep 2021 06:57:00 +0000 Labels:       app=myapp-sts-vol               controller-revision-hash=myapp-sts-vol-c87b47645               statefulset.kubernetes.io/pod-name=myapp-sts-vol-0 Annotations:  cni.projectcalico.org/containerID: 3f1cebedc8fa4bdd2216b5480a754c65fcf9b8167d97c1ba27b203fdf412e2bf               cni.projectcalico.org/podIP: 192.168.9.110/32               cni.projectcalico.org/podIPs: 192.168.9.110/32 Status:       Running IP:           192.168.9.110 IPs:   IP:           192.168.9.110 Controlled By:  StatefulSet/myapp-sts-vol Containers:   myapp:     Container ID:   docker://c2859df65977a15e6b6554d87090a92b41931d1fc901ee0f2aae0f45db8cfee8     Image:          devops2341/go-myweb:alpine     Image ID:       docker-pullable://devops2341/go-myweb@sha256:6ee032cbdc1e9537a2d80f40197836c633b5a03b429504b5b41a814a9185829a     Port:           8080/TCP     Host Port:      0/TCP     State:          Running       Started:      Mon, 06 Sep 2021 06:57:25 +0000     Ready:          True     Restart Count:  0     Environment:    <none>     Mounts:       /data from myapp-data (rw)       /var/run/secrets/kubernetes.io/serviceaccount from default-token-w26v7 (ro) Conditions:   Type              Status   Initialized       True   Ready             True   ContainersReady   True   PodScheduled      True Volumes:   myapp-data:     Type:       PersistentVolumeClaim (a reference to a PersistentVolumeClaim in the same namespace)     ClaimName:  myapp-data-myapp-sts-vol-0     ReadOnly:   false   default-token-w26v7:     Type:        Secret (a volume populated by a Secret)     SecretName:  default-token-w26v7     Optional:    false QoS Class:       BestEffort Node-Selectors:  <none> Tolerations:     node.kubernetes.io/not-ready:NoExecute op=Exists for 300s                  node.kubernetes.io/unreachable:NoExecute op=Exists for 300s Events:   Type     Reason                  Age                    From                     Message   ----     ------                  ----                   ----                     -------   Warning  FailedScheduling        2m23s (x2 over 2m23s)  default-scheduler        0/4 nodes are available: 4 pod has unbound immediate PersistentVolumeClaims.   Normal   Scheduled               2m20s                  default-scheduler        Successfully assigned default/myapp-sts-vol-0 to kube-node1   Normal   SuccessfulAttachVolume  2m20s                  attachdetach-controller  AttachVolume.Attach succeeded for volume "pvc-d0a53c0b-135f-4226-a2e0-89c961bb6a37"   Normal   Pulling                 2m2s                   kubelet                  Pulling image "devops2341/go-myweb:alpine"   Normal   Pulled                  116s                   kubelet                  Successfully pulled image "devops2341/go-myweb:alpine" in 6.108087982s   Normal   Created                 116s                   kubelet                  Created container myapp   Normal   Started                 115s                   kubelet                  Started container myapp vagrant@kube-control1:~/work/20210906$

vagrant@kube-control1:~/work/20210906$ kubectl describe pods myapp-sts-vol-0 Volumes:   myapp-data:     Type:       PersistentVolumeClaim (a reference to a PersistentVolumeClaim in the same namespace)     ClaimName:  myapp-data-myapp-sts-vol-0     ReadOnly:   false   default-token-w26v7:     Type:        Secret (a volume populated by a Secret)     SecretName:  default-token-w26v7     Optional:    false vagrant@kube-control1:~/work/20210906$ kubectl describe pods myapp-sts-vol-1 Volumes:   myapp-data:     Type:       PersistentVolumeClaim (a reference to a PersistentVolumeClaim in the same namespace)     ClaimName:  myapp-data-myapp-sts-vol-1     ReadOnly:   false   default-token-w26v7:     Type:        Secret (a volume populated by a Secret)     SecretName:  default-token-w26v7     Optional:    false vagrant@kube-control1:~/work/20210906$ bash-5.1# host myapp-sts-vol-1.myapp-svc-headless2 myapp-sts-vol-1.myapp-svc-headless2.default.svc.cluster.local has address 192.168.119.177 bash-5.1# host myapp-sts-vol-0.myapp-svc-headless2 myapp-sts-vol-0.myapp-svc-headless2.default.svc.cluster.local has address 192.168.9.79 vagrant@kube-control1:~/work/20210906$ kubectl exec myapp-sts-vol-0 -it -- sh / # cd /data/ /data # ls lost+found           myapp-sts-vol-0.txt  myapp-sts-vol-0_dir /data # cat myapp-sts-vol-0.txt myapp-sts-vol-0 /data # exit vagrant@kube-control1:~/work/20210906$ kubectl exec myapp-sts-vol-1 -it -- sh / # cat /data/myapp-sts-vol-1.txt myapp-sts-vol-1

 

 

 

statefulset은 headless service를 통해 각각의 pods에 직접 접속하는경우 pods에 독립성을 보장해줄 수 있다.

따라서 , (동적 provisioning 된) pvc를 통해 pods가 각각의 독립성을 띈다.