Jusene's Blog

Kubernetes Pod控制器总结

Kubernetes
字数统计: 1.6k阅读时长: 7 min
2018/09/19 Share

Pod

Pod是Kubernetes的最小的原子单位，Pod的创建是分自主式pod资源和Pod控制器创建，首先自主式pod资源清单:

  • apiVersion: kubernetes的master通过apiserver接受jason格式的资源定义，而书写这些资源清单通过yaml的文件定义，apiserver会将这些yaml自动转换jason格式，apiVersion相当于restful的api接口定义:

    1
    ~]# kubectl api-version
    2
    admissionregistration.k8s.io/v1beta1
    3
    apiextensions.k8s.io/v1beta1
    4
    apiregistration.k8s.io/v1
    5
    apiregistration.k8s.io/v1beta1
    6
    apps/v1
    7
    apps/v1beta1
    8
    apps/v1beta2
    9
    authentication.k8s.io/v1
    10
    authentication.k8s.io/v1beta1
    11
    authorization.k8s.io/v1
    12
    authorization.k8s.io/v1beta1
    13
    autoscaling/v1
    14
    autoscaling/v2beta1
    15
    batch/v1
    16
    batch/v1beta1
    17
    certificates.k8s.io/v1beta1
    18
    events.k8s.io/v1beta1
    19
    extensions/v1beta1
    20
    networking.k8s.io/v1
    21
    policy/v1beta1
    22
    rbac.authorization.k8s.io/v1
    23
    rbac.authorization.k8s.io/v1beta1
    24
    scheduling.k8s.io/v1beta1
    25
    storage.k8s.io/v1
    26
    storage.k8s.io/v1beta1
    27
    v1

  • kind: 资源定义

  • metadata: 元数据

  • spec: 期望的状态

  • status: 当前状态，本字段由k8s自己维护

资源的定义字段可以通过以下命令查询:

1
~]# kubectl explain pod.spec

自主式pod

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: pod-daemon
5
  namespace: default
6
  labels:
7
    app: myapp
8
	tier: frontend
9
spec:
10
  containers:
11
    - name: myapp
12
	  image: ikubernetes/myapp:v1
13
	  imagePullPolicy: IfNotPresent
14
	  ports:
15
	  - name: http # 消息型内容
16
	    containerPort: 80
17
	  - name: https
18
	    containerPort: 443
19
    - name: busybox
20
	  image: busybox:latest
21
	  imagePullPolicy: IfNotPresent
22
	  command: ["/bin/sh","-c","sleep 3600"]

创建pod

1
~]# kubectl create -f pod-daemon.yaml
2
~]# kubectl exec -it pod-daemon -c myapp -- /bin/sh # 进入myapp这个容器

资源清单格式:
一级字段:
apiVersion(group/version),kind,metadata(name,namespace,labels,annotations…),spec,status(只读)

Pod资源:

1
spec.containers <[]object>
2
- name: <string>
3
  image: <string>
4
  imagePullPolicy: <string>
5
                   Always,Never,IfNotPresent
6
7
修改镜像的默认应用:
8
  command, args
9
  1. 如果没有command和args，image中的entrypoint和cmd生效
10
  2. 如果存在command没有args，image中的entrypoint和cmd都失效，command生效
11
  3. 如果存在args没有command，image中的entrypoint加args生效
12
  4. 如果存在command和args,images中的entrypoint和cmd失效，command和args生效
13
14
标签：
15
  key=value
16
     key: 字母、数字、_、-、。
17
     value: 可以为空，只能字母或数字开头及结尾，中间可使用
18
  kubectl get pods --show-labels
19
  kubectl get pods -l app --show-labels # 过滤pod对象中app键的pod
20
  kubectl get pods -l app!=dev --show-labels
21
  kubectl get pods -L app,run # 显示对于建的值
22
  kubectl label pods pod-daemon release=dev # 给pod打上标签
23
  kubectl label pods pod-daemon release=qa --overwrite # 给已有标签pod强制打标签
24
25
  等值选择器:
26
      等值关系: =, ==, !=
27
      集合关系: 
28
          KEY in (VALUE1,VALUE2)
29
          KEY notin (VALUE1,VALUE2)
30
          KEY
31
          !KEY
32
33
  许多资源支持内嵌字段定义其使用的标签选择器:
34
                matchLabels: 直接给定键值
35
                matchExpressions: 基于给定的表达式来定义使用标签选择器，{key: "KEY", operator: "OPERATOR", values: ["VAL1", "VAL2"...]}
36
                    操作符(OPERATOR)：In, NotIn: values字段的值必须为非空列表 
37
                                     Exists, NotExists: values字段的值必须为空列表
38
39
  nodeSelector <map[string]string>  
40
            节点标签选择器
41
42
  nodeName: 指定运行那个节点
43
44
  annotaions:
45
            与label不同的地方在于，它不能用于挑选资源对象，仅用于为对象提供‘元数据’。
46
47
  pod的生命周期：
48
            状态：Pending, Running, Faild, Succeeded, Unknown
49
50
            创建Pod: 
51
            Pod生命周期中的重要行为：
52
                初始化容器
53
                容器探测
54
                  liveness # 存活性探测
55
                  readiness # 就绪性探测
56
  restartPolicy:
57
            Always, OnFailure, Never, Default to Always
58
59
  探针类型有三种:
60
            ExecAction、TcpSocketAction、HttpGetAction
61
62
  lifecycle:
63
           postStart:
64
           preStop:
65
        
66
  hostNetwork: 共享主机网络空间

探测

1
apiVersion: v1
2
kind: Pod
3
metadata:
4
  name: liveness-pod
5
  namespace: default
6
spec:
7
  containers:
8
  - name: liveness-exec-container
9
    image: busybox:latest
10
	imagePullPolicy: IfNotPresent
11
	command: ["/bin/sh","-c","touch /tmp/healthy;sleep 3600"]
12
	livenessProbe:
13
	  exec:
14
	    command: ["test","-f","/tmp/healthy"]
15
	  initialDelaySeconds: 3
16
	  periodSeconds: 3
17
	readinessProbe:
18
	  exec:
19
	    command: ["test","-f","/tmp/healthy"]
20
	  initialDelaySeconds: 5
21
	  periodSeconds: 3

存活性探测: 在探测失败的情况下，或根据restartPolicy的规则进行处理，默认Always
就绪性探测: 就绪型探测可以确保服务可用，在service的情况下确保服务可用

Pod控制器

  • ReplicaController
  • ReplicaSet
  • Deployment
  • DaemonSet
  • Job
  • Cronjob
  • StatefulSet

ReplicaSet是ReplicaController是升级版，所以直接使用ReplicaSet来替代ReplicaController。

ReplicaSet

1
apiVersion: apps/v1
2
kind: ReplicaSet
3
metadata:
4
  name: myapp
5
  namespace: default
6
spec:
7
  replicas: 2
8
  selector: 
9
    matchLabels:
10
	  app: myapp
11
	  release: canary
12
  template:
13
    metadata:
14
      name: myapp
15
	  labels:
16
	    app: myapp
17
	    release: canary
18
	    environment: qa
19
    spec:
20
      containers:
21
	  - name: myapp-container
22
	    image: ikubernetes/myapp:v1
23
	    imagePullPolicy: IfNotPresent
24
	    ports;
25
	    - name: http
26
	      containerPort: 80
1
~]# kubectl get rs -o wide
2
NAME                          DESIRED   CURRENT   READY     AGE       CONTAINERS        IMAGES                 SELECTOR
3
myapp                         2         2         2         1d        myapp-container   ikubernetes/myapp:v1   app=myapp,release=canary

ReplicaSet可以根据定义的replicas的pod个数，多减少补，维护块k8s集群中的定义的pod个数。

Deployment

Deployment是定义在ReplicaSet的上一层，通常定义无状态的pod都使用deployment。

1
apiVersion: apps/v1
2
kind: Deployment
3
metadata:
4
  name: myapp-deploy
5
  namespace: default
6
spec:
7
  replicas: 2
8
  selector:
9
    matchLabels:
10
	  app: myapp
11
	  release: canary
12
  template:
13
    metadata:
14
	  labels:
15
		app: myapp
16
		release: canary
17
	spec:
18
	  containers:
19
	  - name: myapp
20
        image: ikubernetes/myapp:v2
21
		imagePullPolicy: IfNotPresent
22
		ports:
23
		- name: http
24
		  containerPort: 80
1
# 以声明式创建
2
~]# kubectl apply -f myapp-deploy.yaml
3
# 补丁
4
~]# kubectl patch deployment myapp-deploy -p '{"spec": {"replicas": 5}}'
5
# 滚动发布
6
~]# kubectl patch deployment myapp-deploy -p '{"spec": {"template": {"spec": {"containers": [{"name":"myapp","image": "ikubernetes/myapp:v3"}]}}}}'
7
# 金丝雀发布,容器中只更新一个容器，就暂停，等稳定后resume更新全部的容器
8
~]# kubectl patch deployment myapp-deploy -p '{"spec": {"strategy": {"rollingUpdate": {"maxSurge": 1, "maxUnavailable": 0}}}}' && kubectl rollout pause deployment myapp-deploy
9
~]# kubectl rollout resume deployment myapp-deploy
10
# 查看发布的版本,默认会维护ReplicaSet的历史版本，默认10个版本
11
~]# kubectl history deploymnet myapp-deploy
12
# 回滚版本
13
~]# kubectl rollout undo deployment myapp-deploy # 回滚到上一个版本
14
~]# kubectl rollout undo deployment myapp-deploy --to-revision=1 # 回滚到指定版本，被回滚的版本更新到最新的版本
15
# 查看回滚的状态
16
~]# kubectl rollout status deployment myapp-deploy

DaemonSet

DaemonSet能让kubernetes每个节点运行同一个pod，每个节点加入kubernetes集群都会自动运行这个pod，适合运行系统性的应用，如日志收集等。

1
apiVersion: apps/v1
2
kind: DaemonSet
3
metadata:
4
  name: myapp-ds
5
  namespace: default
6
spec:
7
  selector:
8
    matchLabels:
9
	  app: filebeat
10
	  release: stable
11
  template:
12
    metadata:
13
	  labels:
14
	    app: filebeat
15
		release: stable
16
    spec:
17
	  containers:
18
	    - name: filebeat
19
		  image: ikubernetes/filebeat:5.6.5-alpine
20
		  imagePullPolicy: IfNotPresent
21
		  env: 
22
		  - name: REDIS_HOST
23
		    value: redis.default.svc.cluster.local
24
		  - name: REDIS_LOG_LEVEL
25
		    value: info

Job

job和rc或者rs的不同在于，job的pod多用于执行一次性任务，执行完成pod后就会停止。

RestartPolicy
job pod的template的RestartPolicy只能指定Never或OnFailure，当job未完成的情况下：

  • 如果RestartPolicy指定Never，则job会在pod出现故障时创建新的pod，且故障pod不会消失。.status.failed加1。
  • 如果RestartPolicy指定OnFailure，则job会在pod出现故障时其内部重启容器，而不是创建pod。.status.failed不变。
1
apiVersion: batch/v1
2
kind: Job
3
metadata:
4
  name: com
5
  namespace: default
6
spec:
7
  template:
8
    metadata:
9
	  name: com
10
	spec:
11
	  containers:
12
	  - name: com
13
	    image: busybox:latest
14
		imagePullPolicy: IfNotPresent
15
		command: ["/bin/sh","-c","sleep 60"]
16
	  restartPolicy: Never
17
	  activeDeadlineSeconds: 70

Cronjob

cronjob的在于计划性job执行。

1
apiVersion: batch/v1beta1
2
kind: CronJob
3
metadata:
4
  name: test
5
  namespace: default
6
spec:
7
  schedule: "* * * * *"
8
  jobTemplate:
9
    spec:
10
	  template:
11
	    spec:
12
		  containers:
13
		  - name: test
14
		    image: busybox:latest
15
			imagePullPolicy: IfNotPresent
16
			command: ["/bin/sh","-c", "echo 'hello world'"]
17
		  restartPolicy: OnFailure

原文作者：Zhang Jusene

原文链接：http://jusene.github.io/2018/09/19/k8s-2/

发表日期：September 19th 2018, 5:07:45 pm

更新日期：November 30th 2019, 1:54:24 am

版权声明：本文采用知识共享署名-非商业性使用 4.0 国际许可协议进行许可

CATALOG
  1. 1. Pod
    1. 1.1. 探测
  2. 2. Pod控制器
    1. 2.1. ReplicaSet
    2. 2.2. Deployment
  3. 3. DaemonSet
  4. 4. Job
  5. 5. Cronjob
Total : 190
2020
2019
2018
2017
python boot troubshooting docker drbd ssh web ELK kubernetes Kubernetes git ansible tools dns corosync consul Django etcd security ftp Go bash bigdata grub heartbeat Java kickstart kvm lvs mongodb mysql mesh nc nfs mogilefs openresty upgrade puppet rabbitmq saltstack rsync screen socat svn tcpdump fastdfs haproxy iscsi keepalived cache openssl vpn optimization prometheus redis RPM cncf supervisord kernel JAVA
python Linux troubshooting Docker docker storage nginx Nosql kubernetes Flask Kubernetes ansible Shell cluster Django Go bigdata boot Java hypervisor mysql mesh tools openresty puppet rabbitmq saltstack optimization cncf varnish tomcat JAVA