K8s（三）Pod资源——pod亲和性与反亲和性，pod重启策略

本文主要介绍了pod资源与pod相关的亲和性，以及pod的重启策略。

心葉493

933人浏览 · 2024-01-16 01:38:01

心葉493 · 2024-01-16 01:38:01 发布

本文主要介绍了pod资源与pod相关的亲和性，以及pod的重启策略

pod亲和性与反亲和性

pod亲和性（podAffinity）有两种 1.podaffinity，即联系比较紧密的pod更倾向于使用同一个区域比如tomcat和nginx这样资源的利用效率更高

2.podunaffinity，即两套完全相同，或两套完全不同功能的服务为了不互相影响容灾效果，或者让服务之间不会互相影响，更倾向于不适用同一个区域

那么如何判断是不是“同一个区域”就非常重要


#查看帮助
kubectl explain pods.spec.affinity.podAffinity
preferredDuringSchedulingIgnoredDuringExecution #软亲和性，尽可能在一起
requiredDuringSchedulingIgnoredDuringExecution  #硬亲和性，一定要在一起

pod亲和性

#硬亲和性
kubectl explain pods.spec.affinity.podAffinity.requiredDuringSchedulingIgnoredDuringExecution
	 labelSelector        <Object>     #以标签为筛选条件，选择一组亲和的pod
   namespaceSelector    <Object>     #以命名空间为筛选条件，选择一组亲和的pod
   namespaces   <[]string>           #确定命名空间的位置
   topologyKey  <string> -required-  #拓扑逻辑键，根据xx判断是否是同一位置

cat > qinhe-pod1.yaml << EOF
apiVersion: v1 
kind: Pod
metadata:
  name: qinhe1
  namespace: default
  labels:
    user: ws
spec:
  containers:
   - name: qinhe1
     image: docker.io/library/nginx
     imagePullPolicy: IfNotPresent
EOF
kubectl apply -f qinhe-pod1.yaml    #定义一个初始的pod，后面的pod可以依次为参照

echo "
apiVersion: v1
kind: Pod
metadata:
  name: qinhe2
  labels:
    app: app1
spec:
  containers:
    - name: qinhe2
      image: docker.io/library/nginx
      imagePullPolicy: IfNotPresent
  affinity:
    podAffinity:  # 和pod亲和性
      requiredDuringSchedulingIgnoredDuringExecution:
        - labelSelector:  # 以标签为筛选条件
            matchExpressions:  # 以表达式进行匹配
             - {key: user, operator: In, values: ["ws"]}
          topologyKey: kubernetes.io/hostname
#带有kubernetes.io/hostname标签相同的被认为是同一个区域，即以主机名区分
#标签的node被认为是统一位置
" > qinhe-pod2.yaml
kubectl apply -f qinhe-pod2.yaml

kubectl get pods -owide #因为hostname node1和node2不同，所以只会调度到node1
NAME     READY   STATUS    RESTARTS   AGE   IP             NODE           NOMINATED NODE   READINESS GATES
qinhe1   1/1     Running   0          68s   10.10.179.9    ws-k8s-node1   <none>           <none>
qinhe2   1/1     Running   0          21s   10.10.179.10   ws-k8s-node1   <none>           <none>

#修改
...
  topologyKey: beta.kubernetes.io/arch
... #node1和node2这两个标签都相同
kubectl delete -f qinhe-pod2.yaml
kubectl apply -f qinhe-pod2.yaml
kubectl get pods -owide #再查看时会发现qinhe2分到了node2
NAME     READY   STATUS    RESTARTS   AGE     IP             NODE           NOMINATED NODE   READINESS GATES
qinhe1   1/1     Running   0          4m55s   10.10.179.9    ws-k8s-node1   <none>           <none>
qinhe2   1/1     Running   0          15s     10.10.234.68   ws-k8s-node2   <none>           <none>

#清理环境
kubectl delete -f qinhe-pod1.yaml
kubectl delete -f qinhe-pod2.yaml

pod反亲和性

kubectl explain pods.spec.affinity.podAntiAffinity
preferredDuringSchedulingIgnoredDuringExecution      <[]Object>
requiredDuringSchedulingIgnoredDuringExecution       <[]Object>

#硬亲和性
#创建qinhe-pod3.yaml
cat > qinhe-pod3.yaml << EOF
apiVersion: v1 
kind: Pod
metadata:
  name: qinhe3
  namespace: default
  labels:
    user: ws
spec:
  containers:
   - name: qinhe3
     image: docker.io/library/nginx
     imagePullPolicy: IfNotPresent
EOF

#创建qinhe-pod4.yaml
echo "
apiVersion: v1
kind: Pod
metadata:
  name: qinhe4
  labels:
    app: app1
spec:
  containers:
    - name: qinhe4
      image: docker.io/library/nginx
      imagePullPolicy: IfNotPresent
  affinity:
    podAntiAffinity:  # 和pod亲和性
      requiredDuringSchedulingIgnoredDuringExecution:
        - labelSelector:  # 以标签为筛选条件
            matchExpressions:  # 以表达式进行匹配
             - {key: user, operator: In, values: ["ws"]}  #表达式user=ws
          topologyKey: kubernetes.io/hostname  #以hostname作为区分是否同个区域
" > qinhe-pod4.yaml
kubectl apply -f qinhe-pod3.yaml
kubectl apply -f qinhe-pod4.yaml
#分配到了不同的node
kubectl get pods -owide
NAME     READY   STATUS    RESTARTS   AGE   IP             NODE           NOMINATED NODE   READINESS GATES
qinhe3   1/1     Running   0          9s    10.10.179.11   ws-k8s-node1   <none>           <none>
qinhe4   1/1     Running   0          8s    10.10.234.70   ws-k8s-node2   <none>           <none>

#修改topologyKey
pod4修改为topologyKey: user
kubectl label nodes ws-k8s-node1 user=xhy
kubectl label nodes ws-k8s-node2 user=xhy
#现在node1和node2都会被pod4识别为同一位置，因为node的label中user值相同

kubectl delete -f qinhe-pod4.yaml
kubectl apply -f qinhe-pod4.yaml
#直接显示离线
kubectl get pods -owide
NAME     READY   STATUS    RESTARTS   AGE     IP             NODE           NOMINATED NODE   READINESS GATES
qinhe3   1/1     Running   0          9m59s   10.10.179.12   ws-k8s-node1   <none>           <none>
qinhe4   0/1     Pending   0          2s      <none>         <none>         <none>           <none>
#查看日志
Warning  FailedScheduling  74s   default-scheduler  0/4 nodes are available: 2 node(s) didn't match pod anti-affinity rules, 2 node(s) had untolerated taint {node-role.kubernetes.io/control-plane: }. preemption: 0/4 nodes are available: 2 No preemption victims found for incoming pod, 2 Preemption is not helpful for scheduling..

#pod反亲和性的软亲和性与node亲和性的软亲和性同理

#清理环境
kubectl label nodes ws-k8s-node1 user-
kubectl label nodes ws-k8s-node2 user-
kubectl delete -f qinhe-pod3.yaml
kubectl delete -f qinhe-pod4.yaml

pod状态与重启策略

参考文档：Pod 的生命周期 | Kubernetes

pod状态

1.pending——挂起
（1）正在创建pod，检查存储、网络、下载镜像等问题
（2）条件不满足，比如硬亲和性，污点等调度条件不满足

2.failed——失败
至少有一个容器因为失败而停止，即非0状态退出

3.unknown——未知
apiserver连不上node节点的kubelet，通常是网络问题

4.Error——错误

5.succeeded——成功
pod所有容器成功终止

6.Unschedulable
pod不能被调度

7.PodScheduled
正在调度中

8.Initialized
pod初始化完成

9.ImagePullBackOff
容器拉取失败

10.evicted
node节点资源不足

11.CrashLoopBackOff
容器曾经启动，但又异常退出了

pod重启策略

当容器异常时，可以通过设置RestartPolicy字段，设置pod重启策略来对pod进行重启等操作

#查看帮助
kubectl explain pod.spec.restartPolicy
KIND:     Pod
VERSION:  v1
FIELD:    restartPolicy <string>
DESCRIPTION:
     Restart policy for all containers within the pod. One of Always, OnFailure,
     Never. Default to Always. More info:
     <https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/#restart-policy>
     Possible enum values:
     - `"Always"`   #只要异常退出，立即自动重启
     - `"Never"`    #不会重启容器
     - `"OnFailure"`#容器错误退出，即退出码不为0时，则自动重启

#测试Always策略，创建always.yaml
cat > always.yaml << EOF
apiVersion: v1
kind: Pod
metadata:
  name: always-pod
  namespace: default
spec:
  restartPolicy: Always
  containers:
   - name: test-pod
     image: docker.io/library/tomcat
     imagePullPolicy: IfNotPresent
EOF
kubectl apply -f always.yaml
kubectl get po #查看状态
NAME         READY   STATUS    RESTARTS   AGE
always-pod   1/1     Running   0          22s
#进入容器去关闭容器
kubectl exec -it always-pod -- /bin/bash
shutdown.sh
#查看当前状态，可以看到always-pod重启计数器为1
kubectl get po
NAME         READY   STATUS    RESTARTS     AGE
always-pod   1/1     Running   1 (5s ago)   70s

#测试never策略，创建never.yaml
cat > never.yaml << EOF
apiVersion: v1
kind: Pod
metadata:
  name: never-pod
  namespace: default
spec:
  restartPolicy: Never
  containers:
   - name: test-pod
     image: docker.io/library/tomcat
     imagePullPolicy: IfNotPresent
EOF
kubectl apply -f never.yaml
kubectl exec -it never-pod -- /bin/bash
shutdown.sh
#不会重启，状态为completed
kubectl get pods | grep never
never-pod    0/1     Completed   0            73s

#测试OnFailure策略，创建onfailure.yaml
cat > onfailure.yaml << EOF
apiVersion: v1
kind: Pod
metadata:
  name: onfailure-pod
  namespace: default
spec:
  restartPolicy: OnFailure
  containers:
   - name: test-pod
     image: docker.io/library/tomcat
     imagePullPolicy: IfNotPresent
EOF
kubectl apply -f onfailure.yaml
#进去后进行异常退出
kubectl exec -it onfailure-pod -- /bin/bash
kill 1
#查看pods状态，已经重启
kubectl get po  | grep onfailure
onfailure-pod   1/1     Running     1 (43s ago)   2m11s
#进入后进行正常退出
kubectl exec -it onfailure-pod -- /bin/bash
shutdown.sh
#查看pods状态，没有重启，进入completed状态
kubectl get po  | grep onfailure
onfailure-pod   0/1     Completed   1             3m58s

#清理环境
kubectl delete -f always.yaml
kubectl delete -f never.yaml
kubectl delete -f onfailure.yaml

K8S/Kubernetes

K8S/Kubernetes社区为您提供最前沿的新闻资讯和知识内容

更多推荐

【深度】阿里巴巴万级规模 K8s 集群全局高可用体系之美

作者 | 韩堂、柘远、沉醉来源 | 阿里巴巴云原生公众号前言台湾作家林清玄在接受记者采访的时候，如此评价自己 30 多年写作生涯：“第一个十年我才华横溢，‘贼光闪现’，令周边黯然失色；第二个十年，我终于‘宝光现形’，不再去抢风头，反而与身边的美丽相得益彰；进入第三个十年，繁华落尽见真醇，我进入了‘醇光初现’的阶段，真正体味到了境界之美”。长夜有穷，真水无香。领略过了 K8s“身在江

K8S/Kubernetes

如何基于 K8s 构建下一代 DevOps 平台？

作者 | 孙健波（天元）导读：当前云原生 DevOps 体系现状如何？面临哪些挑战？如何通过 OAM 解决云原生 DevOps 场景下的诸多问题？云原生开发应用模型 OAM(Open Application Model) 社区核心成员孙健波将为大家一一解答，并分享如何基于 OAM 和 Kubernetes 打造无限能力的下一代 DevOps 平台。什么是 DevOps？为什么基于 Kub