k8s的Pod常见的几种调度形式

1 常见的Pod调度方式：

1 自由调度：默认的kube-scheduler调度

2 定向调度：nodeName, nodeSelector实现

3 亲和性调度：nodeAffinity, PodAffinity, PodAntiAffinity实现

4 污点和容忍调度：taint，tolerations实现

2 实践

2.1 定向调度

nodeName, NodeSelector:

1 首先node结点打标签，disk=ssd

结点打标签

kubectl label nodes k8s-node1 disk=ssd

查看结点标签

kubectl get nodes --show-labels

2 pod模板文件设定NodeSelector的值: disk: ssd

apiVersion: apps/v1

kind: Deployment

metadata:

name: testnginx-deployment

spec:

selector:

matchLabels:

app: testnginx

replicas: 3

template:

metadata:

labels:

app: testnginx

spec:

containers:

name: testnginx

image: nginx:latest

ports:

containerPort: 80

nodeSelector:

disk: ssd

3 预定义标签

kubernetes.io/hostname

2.2 亲和性调度

2.2.1 nodeAffinity

pod与node关系

RequiredDuringSchedulingIgnoredDuringExecution: 必须满足指定规则才能调度pod到Node上

PreferredDuringSchedulingIgnoredDuringExecution：强调优先满足指定规则,软限制

apiVersion: apps/v1

kind: Deployment

metadata:

name: testnginx-deployment

spec:

selector:

matchLabels:

app: testnginx

replicas: 3

template:

metadata:

labels:

app: testnginx

spec:

必须满足指定规则硬限制

复制代码

  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
        nodeSelectorTerms:
        - matchExpressions:
          - key: disk
            operator: In
            values:
            - ssd

尽量优先满足指定规则软限制

复制代码

    affinity:
      nodeAffinity:
        preferredDuringSchedulingIgnoredDuringExecution:
        - weight: 1
          preference:
            matchExpressions:
            - key: disk
              operator: In
              values:
              - ssd
  containers:
  - name: testnginx
    image: nginx:latest
    ports:
    - containerPort: 80

2.2.2 PodAffinity

pod与pod的关系

硬,软亲和性

requiredDuringSchedulingIgnoredDuringExecution 硬性要求，必须要满足条件，保证部署在一起

spec:

affinity:

podAffinity:

requiredDuringSchedulingIgnoredDuringExecution:

labelSelector:

matchExpressions:

-key: app

operator: In

values:

testnginx

topologyKey: "kubernetes.io/hostname"

preferredDuringSchedulingIgnoredDuringExecution 软性要求，尽量满足条件，保证部署在一起

affinity:

podAffinity:

preferredDuringSchedulingIgnoredDuringExecution:

weight: 60

podAffinityTerm:

labelSelector:

matchExpressions:

{key: app, operator: In, values: ["testnginx"]}

topologyKey: kubernetes.io/hostname

weight: 30

podAffinityTerm:

labelSelector:

matchExpressions:

{key: app, operator: In, values: ["backend"]}

topologyKey: kubernetes.io/hostname

2.2.2 PodAntiAffinity

pod与pod的关系

硬，软反亲和性

requiredDuringSchedulingIgnoredDuringExecution 硬性要求，必须满足条件，保证分散部署的最佳实践

#如果节点上的pod标签存在满足app=nginx，则不能部署到节点上

spec:

affinity:

podAntiAffinity:

requiredDuringSchedulingIgnoredDuringExecution:

labelSelector:

matchExpressions:

key: app

operator: In

values:

nginx

topologyKey: "kubernetes.io/hostname"

preferredDuringSchedulingIgnoredDuringExecution 软性要求，可以不完全满足，可同一node跑多个副本

软性要求

如果节点上的pod标签存在满足app=nginx，也可以部署到节点上，尽可能先部署到其它节点，如果没有满足也可以部署到此节点（大概是这么理解吧）

复制代码

spec:
  affinity:
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
        - labelSelector:
            matchExpressions:
            - key: app
              operator: In
              values:
              - nginx
          topologyKey: "kubernetes.io/hostname"

2.3 污点容忍调度

污点（Taint）是应用在节点之上的，为了排斥pod 所存在的

容忍度（Toleration）是应用于 Pod 上的，允许（但并不要求）Pod 调度到带有与之匹配的污点的节点上。

结点亲和性，是pod的一种属性(偏好或硬件要求)，它使pod被吸引到一类特点的结点

Taint则相反，它使结点能够排斥一类特点的Pod，Taint和Toleration相互配合使用

2.3.1 污点 taint

设置污点

kubectl taint nodes node1 xtz=value1:NoSchedule

去除污点

kubectl taint nodes node1 xtz:NoSchedule-

#节点说明中，查找 Taints 字段

kubectl describe node node-name

三种调度策略：

PreferNoSchedule(软策略，尽量不调度到污点结点)

NoExecute(不会将 Pod 调度到具有该污点的 Node 上，同时会将 Node 上已经存在的 Pod驱逐出去）

NoSchedule(不会将Pod调度到具有该污点的Node上)

2.3.2 容忍度tolerations

spec:

containers:

name: nginx
image: nginx
imagePullPolicy: IfNotPresent
tolerations:
key: "key1"
operator: "Equal"
value: "value1"
effect: "NoSchedule"
3 参考资料
https://blog.csdn.net/weixin_45310323/article/details/130718456
https://zhuanlan.zhihu.com/p/188514195
https://blog.csdn.net/qq_20042935/article/details/128442108
https://www.cnblogs.com/ygbh/p/17290915.html#_label1
https://blog.csdn.net/manwufeilong/article/details/124821748
https://www.cnblogs.com/ygbh/p/17290915.html#_lab2_1_1

k8s的Pod常见的几种调度形式

结点打标签

查看结点标签

必须满足指定规则 硬限制

尽量优先满足指定规则 软限制

软性要求

如果节点上的pod标签存在满足app=nginx，也可以部署到节点上，尽可能先部署到其它节点，如果没有满足也可以部署到此节点（大概是这么理解吧）

设置污点

去除污点

必须满足指定规则硬限制

尽量优先满足指定规则软限制