pod中影响调度的主要属性

资源限制limit、request

容器资源限制：

• resources.limits.cpu
• resources.limit.memory

容器使用的最小资源要求，作为容器调度时资源分配的依据：

• resources.requests.cpu
• resources.requests.memory

k8s会根据request的值五去查找有足够资源的node来调度pod

CPU单位：可以写m也可以写浮点数，例如0.5=500m,1=1000m

apiVersion: v1
kind: pod
metadata:
  name: my-pod
spec:
 containers:
 - name: web
   image: nginx
   resoures:
     request:
       memory: "64Mi"
       cpu: "250m"
     limits:
       memory: "128Mi"
       cpu: "500m"

limits和request的配置简单，但是设置这些值是有讲究的

• limits的值，不能超过节点的配置
• request的值不能超过limits的值
• request是最小资源需求，是一种预留性质的，也就是不是写多少，就会占用宿主机多少资源
• 如果request设置过高就会导致可用节点少，资源空闲，反之过少就会导致pod多，资源饱和
• 当request的值在节点不能满足时，pod处于pending状态

查看一个节点的资源占用情况,会显示每个pod的request和limit以及总和

Namespace                   Name                                      CPU Requests  CPU Limits  Memory Requests  Memory Limits  Age
  ---------                   ----                                      ------------  ----------  ---------------  -------------  ---
  kube-system                 calico-kube-controllers-d4bfdcb9-vss4q    0 (0%)        0 (0%)      0 (0%)           0 (0%)         14d
  kube-system                 calico-node-7pwmf                         250m (8%)     0 (0%)      0 (0%)           0 (0%)         14d
  kube-system                 coredns-6d8c4cb4d-2sqgp                   100m (3%)     0 (0%)      70Mi (2%)        170Mi (5%)     14d
  kube-system                 coredns-6d8c4cb4d-5q9ks                   100m (3%)     0 (0%)      70Mi (2%)        170Mi (5%)     14d
  kube-system                 etcd-k8s-master                           100m (3%)     0 (0%)      100Mi (3%)       0 (0%)         14d
  kube-system                 kube-apiserver-k8s-master                 250m (8%)     0 (0%)      0 (0%)           0 (0%)         14d
  kube-system                 kube-controller-manager-k8s-master        200m (6%)     0 (0%)      0 (0%)           0 (0%)         14d
  kube-system                 kube-proxy-6m547                          0 (0%)        0 (0%)      0 (0%)           0 (0%)         14d
  kube-system                 kube-scheduler-k8s-master                 100m (3%)     0 (0%)      0 (0%)           0 (0%)         14d
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  Resource           Requests     Limits
  --------           --------     ------
  cpu                1100m (36%)  0 (0%)
  memory             240Mi (7%)   340Mi (10%)
  ephemeral-storage  0 (0%)       0 (0%)
  hugepages-1Gi      0 (0%)       0 (0%)
  hugepages-2Mi      0 (0%)       0 (0%)
  hugepages-32Mi     0 (0%)       0 (0%)
  hugepages-64Ki     0 (0%)       0 (0%)

NodeSelector: 将pod和node进行绑定

用于将pod调度到匹配Label标签的节点上，如果没有匹配的标签就会调度失败

作用：

• 约束pod到特定的节点上运行
• 完全匹配标签，一旦匹配不到就调度失败

应用场景：

• 专用节点：根据业务线将node分组管理（比如：A业务线的pod需要全部调度到有 test:aa的节点上去）
• 匹配特殊的硬件：部分Node有ssd盘

这样的配置意思是这个pod永远只能在携带了disktype:ssd标签的node上运行，否则将调度失败。

NodeName，一旦pod的这个字段被赋值，k8s就会认为这个pod已经调度成功，NodeName的值就是node节点名称。这个字段一般由调度器负责设置，测试的时候也可以人为设置。

示例：确保pod分配到有ssd硬盘的节点上

1、给节点添加标签

kubectl label nodes <node_name> <label-key>=<lable-name>
kubectl label nodes k8s-node1 disktype=ssd
#  kubectl get nodes --show-labels 查看节点标签

2、添加nodeSelect字段到pod中

apiVersion: v1
kind: Pod
metadata:
  name: test1
spec:
  nodeSelector:
    disktype: "ssd"
  containers:
  - name: nginx
    image: nginx:v1.18

3、查看pod节点分配

kubectl get pod -A -o wide
# 删除节点标签 kubectl label node k8s-node1 <lable-key>-
 kubectl label nodes k8s-node1 disktype-

nodeAffinity 节点亲和

节点亲和类似于nodeSelector，也是用于根据node上的标签把pod调度到指定节点上

相比nodeSelector:

• 有更多的匹配逻辑选择，不仅仅是完全匹配

In、NotIn、Exists、DoesNotExist、Gt、Lt

• 调度策略分为软策略和硬策略

硬策略（required）: 必须满足

软策略（preferred）: 尝试满足，但不保证

apiVersion: v1
kind: Pod
metadata:
  name: nginx1
spec:
  affinity:
    nodeAffinity:
      requiredDuringSchedulingIgnoredDuringExecution: # 硬策略
        nodeSelectorTerms:
        - matchExpressions:
          - key: disktype # key类型
            operator: In # 策略逻辑
            values: #  值
            - ssd         
  containers:
  - name: nginx
    image: nginx

apiVersion: v1
kind: Pod
metadata:
  name: nginx55
spec:
  affinity:
    nodeAffinity:
      preferredDuringSchedulingIgnoredDuringExecution: # 软策略
      - weight: 1 # 权重
        preference:
          matchExpressions:
          - key: disktype
            operator: In
            values:
            - ssd88         
  containers:
  - name: nginx
    image: nginx