kubernete编排技术二：deployment

gq_design

发布于 2022-5-1 09:33

浏览

0收藏

作者 | 朱晋君
来源 | 君哥聊技术（ID：gh_1f109b82d301）

kubernete中的控制器模式，是指用一种对象来控制另一种对象，这个控制器是由组件kube-controller-manager来实现的。deployment就是一种控制器，用来控制pod的创建、水平扩展收缩、滚动更新。deployment是kubernete中一个对无状态pod的控制器，本文来介绍一下。

deployment简介

先回顾一下我们之前定义的yaml文件，这个文件定义了2个pod副本(replicas: 2),这个deployment的编排逻辑就是带app: springboot-mybatis这个label的数量只能有2个，超过就需要删除，少了就需要创建。

apiVersion: apps/v1
kind: Deployment
metadata:
  name: springboot-mybatis-deployment
spec:
  selector:
    matchLabels:
      app: springboot-mybatis
  replicas: 2
  template:
    metadata:
      labels:
        app: springboot-mybatis
    spec:
      containers:
      - name: spingboot-mybatis
        imagePullPolicy: IfNotPresent
        image: zjj2006forever/springboot-mybatis:1.2
        ports:
        - containerPort: 83001.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.

上面yaml文件中的template，被称作pod模板，而template上面的代码部分，就是控制器的定义。这儿的控制器其实就是上面的控制器部分对下面pod模板的控制。创建时，控制器会从etcd查找label是springboot-mybatis的pod个数，因为是0，所以会创建2个;如果运行过程中，有一个pod挂了，会再创建一个。

pod水平扩展和收缩

上一节的yaml文件定义了一个2个pod的deployment控制器，这样会生成2个pod。在我们实际工作中，有时会遇到突增流量，比如促销活动之类的，这时候我们可以用增加pod来应对。最直接的方式就是修改replicas，比如改为replicas: 4。但是生成环境中，这样效率太低了，我们可以执行如下命令：

kubectl scale deployment springboot-mybatis-deployment --replicas=41.

之后我们查看集群上的pod，这2个AGE=10s的pod就是后来扩展出来的。

[root@master k8s]# kubectl get pods
NAME                                             READY   STATUS    RESTARTS   AGE
springboot-mybatis-deployment-5b78f66997-d5fk9   1/1     Running   0          61s
springboot-mybatis-deployment-5b78f66997-hcmm4   1/1     Running   0          10s
springboot-mybatis-deployment-5b78f66997-k5xw8   1/1     Running   0          61s
springboot-mybatis-deployment-5b78f66997-pvfdv   1/1     Running   0          11s1.
2.
3.
4.
5.
6.

我们查看其中一个新pod的状态，跟之前的pod创建过程完全一样。

[root@master k8s]# kubectl describe pod springboot-mybatis-deployment-5b78f66997-hcmm4
Name:         springboot-mybatis-deployment-5b78f66997-hcmm4
Namespace:    default
Priority:     0
Node:         worker1/192.168.59.141
Start Time:   Wed, 08 Jul 2020 06:01:08 -0400
Labels:       app=springboot-mybatis
              pod-template-hash=5b78f66997
Annotations:  <none>
Status:       Running
IP:           10.244.1.25
IPs:
  IP:           10.244.1.25
Controlled By:  ReplicaSet/springboot-mybatis-deployment-5b78f66997
Containers:
  spingboot-mybatis:
    Container ID:   docker://ef433a219549b32e1d01899ea6ac581c1bfffa7d01e1c165647be52f9f79ff5b
    Image:          zjj2006forever/springboot-mybatis:1.2
    Image ID:       docker-pullable://zjj2006forever/springboot-mybatis@sha256:bf43bc9d1d4bdb33d82a206282c8f82be3679847d1011806b72e3634ebe67564
    Port:           8300/TCP
    Host Port:      0/TCP
    State:          Running
      Started:      Wed, 08 Jul 2020 06:01:11 -0400
    Ready:          True
    Restart Count:  0
    Environment:    <none>
    Mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from default-token-qk5nc (ro)
Conditions:
  Type              Status
  Initialized       True 
  Ready             True 
  ContainersReady   True 
  PodScheduled      True 
Volumes:
  default-token-qk5nc:
    Type:        Secret (a volume populated by a Secret)
    SecretName:  default-token-qk5nc
    Optional:    false
QoS Class:       BestEffort
Node-Selectors:  <none>
Tolerations:     node.kubernetes.io/not-ready:NoExecute for 300s
                 node.kubernetes.io/unreachable:NoExecute for 300s
Events:
  Type    Reason     Age        From               Message
  ----    ------     ----       ----               -------
  Normal  Scheduled  <unknown>  default-scheduler  Successfully assigned default/springboot-mybatis-deployment-5b78f66997-hcmm4 to worker1
  Normal  Pulled     107s       kubelet, worker1   Container image "zjj2006forever/springboot-mybatis:1.2" already present on machine
  Normal  Created    106s       kubelet, worker1   Created container spingboot-mybatis
  Normal  Started    106s       kubelet, worker1   Started container spingboot-mybatis1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.

查看replicas

[root@master k8s]# kubectl get rs
NAME                                       DESIRED   CURRENT   READY   AGE
springboot-mybatis-deployment-5b78f66997   4         4         4       3m59s1.
2.
3.

检测一下应用运行正常

[root@worker1 ~]# curl http://10.244.1.24:18082/actuator/info
{"app":{"name":"springboot-mybatis","description":"two-datasource-test","version":"1.0","spring-boot-version":"2.1.6"}}1.
2.

如果流量降下来了，我们可以收缩pod数量，执行命令

kubectl scale deployment nginx-deployment --replicas=21.

查看pod数量，后来扩展的2个pod被删除了

[root@master k8s]# kubectl get pods
NAME                                             READY   STATUS    RESTARTS   AGE
springboot-mybatis-deployment-5b78f66997-d5fk9   1/1     Running   0          11m
springboot-mybatis-deployment-5b78f66997-k5xw8   1/1     Running   0          11m1.
2.
3.
4.

可以看出，这种控制器模式，是由deployment控制ReplicaSet，ReplicaSet控制pod。

滚动更新

在实际开发工作中，升级服务上线是经常的事情。以本文中的应用为例，之前的deployment使用的镜像是zjj2006forever/springboot-mybatis:1.2，如果我们升级镜像版本到1.3，需要怎么操作呢?

注:镜像的升级版本我已经打好并且发布到dockerhub，感兴趣的同学自行拉取测试。

这个升级操作我们可以直接修改yaml文件，也可以执行如下命令进行修改，这个命令的原理是直接编辑etcd里面的springboot-mybatis-deployment控制器对象，可以跟编辑文件一样进行编辑

kubectl edit deployment/springboot-mybatis-deployment1.

编辑完成保存后，就会触发滚动更新。执行下面命令查看滚动更新的日志。

[root@master ~]# kubectl rollout status deployment/springboot-mybatis-deployment
Waiting for deployment "springboot-mybatis-deployment" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "springboot-mybatis-deployment" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "springboot-mybatis-deployment" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "springboot-mybatis-deployment" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "springboot-mybatis-deployment" rollout to finish: 1 old replicas are pending termination...
deployment "springboot-mybatis-deployment" successfully rolled out1.
2.
3.
4.
5.
6.
7.

这时我们再次查看pod，发现新的pod已经创建出来了

[root@master k8s]# kubectl get pods
NAME                                             READY   STATUS    RESTARTS   AGE
springboot-mybatis-deployment-69787d89c4-8kplf   1/1     Running   0          30s
springboot-mybatis-deployment-69787d89c4-hv4zl   1/1     Running   0          3m11s1.
2.
3.
4.

从下面springboot-mybatis-deployment的描述Events中可以看出滚动更新的过程

[root@master ~]# kubectl describe deployment springboot-mybatis-deployment
Name:                   springboot-mybatis-deployment
Namespace:              default
CreationTimestamp:      Wed, 08 Jul 2020 06:00:17 -0400
Labels:                 <none>
Annotations:            deployment.kubernetes.io/revision: 2
                        kubectl.kubernetes.io/last-applied-configuration:
                          {"apiVersion":"apps/v1","kind":"Deployment","metadata":{"annotations":{},"name":"springboot-mybatis-deployment","namespace":"default"},"sp...
Selector:               app=springboot-mybatis
Replicas:               2 desired | 2 updated | 2 total | 2 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=springboot-mybatis
  Containers:
   spingboot-mybatis:
    Image:        zjj2006forever/springboot-mybatis:1.3
    Port:         8300/TCP
    Host Port:    0/TCP
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    NewReplicaSetAvailable
OldReplicaSets:  <none>
NewReplicaSet:   springboot-mybatis-deployment-69787d89c4 (2/2 replicas created)
Events:
  Type    Reason             Age    From                   Message
  ----    ------             ----   ----                   -------
  Normal  ScalingReplicaSet  30m    deployment-controller  Scaled up replica set springboot-mybatis-deployment-5b78f66997 to 2
  Normal  ScalingReplicaSet  29m    deployment-controller  Scaled up replica set springboot-mybatis-deployment-5b78f66997 to 4
  Normal  ScalingReplicaSet  19m    deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 2
  Normal  ScalingReplicaSet  6m18s  deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 1
  Normal  ScalingReplicaSet  3m37s  deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 1
  Normal  ScalingReplicaSet  3m37s  deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 2
  Normal  ScalingReplicaSet  3m32s  deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 01.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.

上面event中最后4行就是滚动更新的过程。很明显，滚动更新过程是创建一个新的pod，删除一个旧的pod，直到新的pod全部创建，旧的pod的全部删除。可以"滚动更新"这个词非常形象。滚动更新有如下优势：

1.新pod启动失败整个过程就会停止，剩下的旧pod依然可以提供服务。

2.为了保证服务不中断，deployment控制器会保证一次删除pod的比例和一次创建的pod的比例，默认都是yaml中replicas的25%。

3.这2个比例也可以通过参数来配置，看下面的yaml。maxSurge指每次滚动过程中可以创建多少个pod，maxUnavailable指一次更新中可以删除几个pod，这2个参数值也可以用百分数表示，指的是滚动过程中一次创建或删除的pod数量占replicas配置的pod的比例。

apiVersion: apps/v1
kind: Deployment
metadata:
  name: springboot-mybatis-deployment
spec:
  selector:
    matchLabels:
      app: springboot-mybatis
  replicas: 2
  template:
    metadata:
      labels:
        app: springboot-mybatis
    spec:
      containers:
      - name: spingboot-mybatis
        imagePullPolicy: IfNotPresent
        image: zjj2006forever/springboot-mybatis:1.3
        ports:
        - containerPort: 8300
    strategy:
      type: RollingUpdate
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 11.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.

失败回滚

如果我们升级应用的过程中，pod启动失败，这时候我们为了不中断服务，必须回滚。比如之前的yaml文件中，我们定义的springboot-mybatis镜像版本是1.3，如果我们改成1.4版本，dockerhub上并没有这个版本，拉取失败导致pod启动失败。修改yaml文件中镜像版本到1.4后，执行更新命令

kubectl apply -f springboot-mybatis.yaml1.

这时deployment就会触发滚动更新，但是这时因为镜像拉取失败，pod创建失败，如下结果中第4行

[root@master ~]# kubectl get rs
NAME                                       DESIRED   CURRENT   READY   AGE
springboot-mybatis-deployment-5b78f66997   0         0         0       15h
springboot-mybatis-deployment-69787d89c4   2         2         2       14h
springboot-mybatis-deployment-7d6fc78b87   1         1         0       80s1.
2.
3.
4.
5.

我们再查看deployment的状态，如下，镜像版本是1.4，这时我们可以触发回滚操作。

[root@master ~]# kubectl describe deployment springboot-mybatis-deployment
Name:                   springboot-mybatis-deployment
Namespace:              default
CreationTimestamp:      Wed, 08 Jul 2020 06:00:17 -0400
Labels:                 <none>
Annotations:            deployment.kubernetes.io/revision: 3
                        kubectl.kubernetes.io/last-applied-configuration:
                          {"apiVersion":"apps/v1","kind":"Deployment","metadata":{"annotations":{},"name":"springboot-mybatis-deployment","namespace":"default"},"sp...
Selector:               app=springboot-mybatis
Replicas:               2 desired | 1 updated | 3 total | 2 available | 1 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=springboot-mybatis
  Containers:
   spingboot-mybatis:
    Image:        zjj2006forever/springboot-mybatis:1.4
    Port:         8300/TCP
    Host Port:    0/TCP
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    ReplicaSetUpdated
OldReplicaSets:  springboot-mybatis-deployment-69787d89c4 (2/2 replicas created)
NewReplicaSet:   springboot-mybatis-deployment-7d6fc78b87 (1/1 replicas created)
Events:
  Type    Reason             Age   From                   Message
  ----    ------             ----  ----                   -------
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 1
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 1
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 2
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 0
  Normal  ScalingReplicaSet  3m3s  deployment-controller  Scaled up replica set springboot-mybatis-deployment-7d6fc78b87 to 11.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.

下面命令可以查看deployment的升级历史

[root@master k8s]# kubectl rollout history deployment/springboot-mybatis-deployment
deployment.apps/springboot-mybatis-deployment 
REVISION  CHANGE-CAUSE
1         <none>
2         <none>
3         <none>1.
2.
3.
4.
5.
6.

从上面的deployment的describe中我们可以看出目前revision是3(deployment.kubernetes.io/revision: 3),我们应该恢复到2，如下命令

kubectl rollout undo deployment/springboot-mybatis-deployment --to-revision=21.

执行之后，再看deployment状态，镜像版本已经恢复到zjj2006forever/springboot-mybatis:1.3

[root@master k8s]# kubectl describe deployment
Name:                   springboot-mybatis-deployment
Namespace:              default
CreationTimestamp:      Wed, 08 Jul 2020 06:00:17 -0400
Labels:                 <none>
Annotations:            deployment.kubernetes.io/revision: 4
                        kubectl.kubernetes.io/last-applied-configuration:
                          {"apiVersion":"apps/v1","kind":"Deployment","metadata":{"annotations":{},"name":"springboot-mybatis-deployment","namespace":"default"},"sp...
Selector:               app=springboot-mybatis
Replicas:               2 desired | 2 updated | 2 total | 2 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=springboot-mybatis
  Containers:
   spingboot-mybatis:
    Image:        zjj2006forever/springboot-mybatis:1.3
    Port:         8300/TCP
    Host Port:    0/TCP
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    NewReplicaSetAvailable
OldReplicaSets:  <none>
NewReplicaSet:   springboot-mybatis-deployment-69787d89c4 (2/2 replicas created)
Events:
  Type    Reason             Age   From                   Message
  ----    ------             ----  ----                   -------
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 1
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 1
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled up replica set springboot-mybatis-deployment-69787d89c4 to 2
  Normal  ScalingReplicaSet  14h   deployment-controller  Scaled down replica set springboot-mybatis-deployment-5b78f66997 to 0
  Normal  ScalingReplicaSet  15m   deployment-controller  Scaled up replica set springboot-mybatis-deployment-7d6fc78b87 to 1
  Normal  ScalingReplicaSet  35s   deployment-controller  Scaled down replica set springboot-mybatis-deployment-7d6fc78b87 to 01.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.

最后，从replicas的结果中我们可以看出，每次无论升级成功失败都会有一个replicas创建出来，为了限制保留的历史版本个数，可以通过参数spec.revisionHistoryLimit来进行设置。

[root@master k8s]# kubectl get rs
NAME                                       DESIRED   CURRENT   READY   AGE
springboot-mybatis-deployment-5b78f66997   0         0         0       15h
springboot-mybatis-deployment-69787d89c4   2         2         2       14h
springboot-mybatis-deployment-7d6fc78b87   0         0         0       14m1.
2.
3.
4.
5.

总结

deployment是kubernete中非常重要的一个编排技术，不过它有一个局限，只能编排无状态的pod，比如一个应用的集群，集群中的节点都是平等的，不存在主备关系。但是我们日常的开发中，好多pod是有状态的，比如一个部署在kubernete中的redis集群，会存在主从节点，这样启动的时候pod就存在顺序，deployment就不能满足了。

分类

云原生

标签

kubernete

deployment

51CTO

51CTO博客

51CTO学堂

kubernete编排技术二：deployment

订阅鸿蒙技术特刊，精选内容抢先看