Linux - K8S - 调度策略 - Pod策略之亲和和反亲和性

?
1
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
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
## Pod调度 - 亲和podAffinity
 
### 硬亲和requiredDuringSchedulingIgnoredDuringExecution
 
```sh
[15:25:06 root@master1 scheduler]#cat 07-scheduler-pod-affinity-base.yaml
---
apiVersion: v1
kind: Pod
metadata:
  name: pod-dev
  labels:
    env: dev
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  nodeSelector:
    env: dev
---
apiVersion: v1
kind: Pod
metadata:
  name: pod-test
  labels:
    env: test
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  nodeSelector:
    env: test
[15:25:18 root@master1 scheduler]#kubectl apply -f 07-scheduler-pod-affinity-base.yaml
pod/pod-dev created
pod/pod-test created
[15:25:27 root@master1 scheduler]#kubectl get pod -o wide
NAME       READY   STATUS    RESTARTS   AGE   IP            NODE               NOMINATED NODE   READINESS GATES
pod-dev    1/1     Running   0          5s    10.244.3.10   node1.noisedu.cn   <none>           <none>
pod-test   1/1     Running   0          5s    10.244.4.9    node2.noisedu.cn   <none>           <none>
 
# 此时建立了两个pod分别有各自的标签，下面根据pod的标签创建基于硬亲和的pod
[15:26:06 root@master1 scheduler]#cat 08-scheduler-pod-affinity-required.yaml
apiVersion: v1
kind: Pod
metadata:
  name: pod-affinity
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  affinity:
    podAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - key: env
            operator: In
            values:
            - test
        namespaces: ["default"]
        topologyKey: kubernetes.io/hostname
 
[15:25:32 root@master1 scheduler]#kubectl apply -f 08-scheduler-pod-affinity-required.yaml
pod/pod-affinity created
[15:25:40 root@master1 scheduler]#kubectl get pod -o wide
NAME           READY   STATUS    RESTARTS   AGE   IP            NODE               NOMINATED NODE   READINESS GATES
pod-affinity   1/1     Running   0          4s    10.244.4.10   node2.noisedu.cn   <none>           <none>
pod-dev        1/1     Running   0          17s   10.244.3.10   node1.noisedu.cn   <none>           <none>
pod-test       1/1     Running   0          17s   10.244.4.9    node2.noisedu.cn   <none>           <none>
[15:25:44 root@master1 scheduler]#kubectl get pod -o wide --show-labels
NAME           READY   STATUS    RESTARTS   AGE   IP            NODE               NOMINATED NODE   READINESS GATES   LABELS
pod-affinity   1/1     Running   0          26s   10.244.4.10   node2.noisedu.cn   <none>           <none>            <none>
pod-dev        1/1     Running   0          39s   10.244.3.10   node1.noisedu.cn   <none>           <none>            env=dev
pod-test       1/1     Running   0          39s   10.244.4.9    node2.noisedu.cn   <none>           <none>            env=test
 
 
 
 
 
```
 
### 软亲和preferredDuringSchedulingIgnoredDuringExecution
 
```sh
# 多满足条件，选择权重高的node
[15:30:53 root@master1 scheduler]#cat 09-scheduler-pod-affinity-preferred.yaml
apiVersion: v1
kind: Pod
metadata:
  name: pod-affinity
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  affinity:
    podAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 60
        podAffinityTerm:
          labelSelector:
            matchExpressions:
            - {key: env, operator: In, values: ["dev"]}
          topologyKey: kubernetes.io/hostname
      - weight: 30
        podAffinityTerm:
          labelSelector:
            matchExpressions:
            - {key: env, operator: In, values: ["test"]}
          topologyKey: kubernetes.io/hostname
[15:31:26 root@master1 scheduler]#kubectl apply -f 09-scheduler-pod-affinity-preferred.yaml
pod/pod-affinity created
[15:32:32 root@master1 scheduler]#kubectl get pod -o wide
NAME           READY   STATUS    RESTARTS   AGE     IP            NODE               NOMINATED NODE   READINESS GATES
pod-affinity   1/1     Running   0          9s      10.244.3.11   node1.noisedu.cn   <none>           <none>
pod-dev        1/1     Running   0          7m14s   10.244.3.10   node1.noisedu.cn   <none>           <none>
pod-test       1/1     Running   0          7m14s   10.244.4.9    node2.noisedu.cn   <none>           <none>
 
# 这种方式，一般会适用于一些服务和后台存储必须放在一起
```
 
## Pod调度 - 反亲和podAntiAffinity
 
```sh
[15:45:56 root@master1 scheduler]#cat 12-scheduler-pod-anaffinity-required.yaml
apiVersion: v1
kind: Pod
metadata:
  name: pod-antiaffinity
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  affinity:
    podAntiAffinity:
      requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
          - {key: env, operator: In, values: ["dev"]}
        topologyKey: kubernetes.io/hostname
[15:46:08 root@master1 scheduler]#kubectl apply -f 12-scheduler-pod-anaffinity-required.yaml
pod/pod-antiaffinity created
[15:46:31 root@master1 scheduler]#kubectl get pod -o wide
NAME               READY   STATUS    RESTARTS   AGE   IP            NODE               NOMINATED NODE   READINESS GATES
pod-antiaffinity   1/1     Running   0          5s    10.244.4.11   node2.noisedu.cn   <none>           <none>
pod-dev            1/1     Running   0          21m   10.244.3.10   node1.noisedu.cn   <none>           <none>
pod-test           1/1     Running   0          21m   10.244.4.9    node2.noisedu.cn   <none>           <none>
[15:46:36 root@master1 scheduler]#kubectl get pod -o wide --show-labels
NAME               READY   STATUS    RESTARTS   AGE   IP            NODE               NOMINATED NODE   READINESS GATES   LABELS
pod-antiaffinity   1/1     Running   0          24s   10.244.4.11   node2.noisedu.cn   <none>           <none>            <none>
pod-dev            1/1     Running   0          21m   10.244.3.10   node1.noisedu.cn   <none>           <none>            env=dev
pod-test           1/1     Running   0          21m   10.244.4.9    node2.noisedu.cn   <none>           <none>            env=test
 
# 因为是反亲和性，所以会调度到pod-test的相应节点
```

　　

?
1
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
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
# 软反亲和
[22:09:48 root@master1 scheduler]#cat 13-scheduler-pod-anaffinity-preferred.yaml
apiVersion: v1
kind: Pod
metadata:
  name: pod-antiaffinity
spec:
  containers:
  - name: pod-test
    image: 10.0.0.55:80/mykubernetes/pod_test:v0.1
    imagePullPolicy: IfNotPresent
  affinity:
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
        - weight: 100
          podAffinityTerm:
            labelSelector:
              matchExpressions:
              - {key: env, operator: In, values: ["dev"]}
            topologyKey: kubernetes.io/hostname
        - weight: 50
          podAffinityTerm:
            labelSelector:
              matchExpressions:
              - {key: env, operator: In, values: ["test"]}
            topologyKey: kubernetes.io/hostname
[22:09:51 root@master1 scheduler]#kubectl apply -f 13-scheduler-pod-anaffinity-preferred.yaml
pod/pod-antiaffinity created
[22:10:03 root@master1 scheduler]#kubectl get pod -o wide
NAME               READY   STATUS              RESTARTS   AGE   IP           NODE               NOMINATED NODE   READINESS GATES
pod-antiaffinity   0/1     ContainerCreating   0          5s    <none>       node2.noisedu.cn   <none>           <none>
pod-dev            1/1     Running             0          21s   10.244.3.3   node1.noisedu.cn   <none>           <none>
pod-test           1/1     Running             0          21s   10.244.4.3   node2.noisedu.cn   <none>           <none>
[22:10:08 root@master1 scheduler]#kubectl get pod -o wide
NAME               READY   STATUS    RESTARTS   AGE   IP           NODE               NOMINATED NODE   READINESS GATES
pod-antiaffinity   1/1     Running   0          12s   10.244.4.4   node2.noisedu.cn   <none>           <none>
pod-dev            1/1     Running   0          28s   10.244.3.3   node1.noisedu.cn   <none>           <none>
pod-test           1/1     Running   0          28s   10.244.4.3   node2.noisedu.cn   <none>           <none>
[22:10:15 root@master1 scheduler]#kubectl describe pod pod-antiaffinity
Name:         pod-antiaffinity
Namespace:    default
Priority:     0
Node:         node2.noisedu.cn/10.0.0.54
Start Time:   Sun, 16 Jan 2022 22:10:03 +0800
Labels:       <none>
Annotations:  cni.projectcalico.org/containerID: 62ba17bfd69068d6c93cb2ca894749e93af745193ecb0f7bddc125847567894f
              cni.projectcalico.org/podIP: 10.244.4.4/32
              cni.projectcalico.org/podIPs: 10.244.4.4/32
Status:       Running
IP:           10.244.4.4
IPs:
  IP:  10.244.4.4
Containers:
  pod-test:
    Container ID:   docker://3b8321e664753a0c54fbc74b903cd7c0614e03d44e33618b3998b3d119b9fbf1
    Image:          10.0.0.55:80/mykubernetes/pod_test:v0.1
    Image ID:       docker-pullable://10.0.0.55:80/mykubernetes/pod_test@sha256:54402cda2ef15f45e4aafe98a5e56d4de076e3d4100c2a1bf1b780c787372fed
    Port:           <none>
    Host Port:      <none>
    State:          Running
      Started:      Sun, 16 Jan 2022 22:10:08 +0800
    Ready:          True
    Restart Count:  0
    Environment:    <none>
    Mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-vclbc (ro)
Conditions:
  Type              Status
  Initialized       True
  Ready             True
  ContainersReady   True
  PodScheduled      True
Volumes:
  kube-api-access-vclbc:
    Type:                    Projected (a volume that contains injected data from multiple sources)
    TokenExpirationSeconds:  3607
    ConfigMapName:           kube-root-ca.crt
    ConfigMapOptional:       <nil>
    DownwardAPI:             true
QoS Class:                   BestEffort
Node-Selectors:              <none>
Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
Events:
  Type    Reason     Age   From               Message
  ----    ------     ----  ----               -------
  Normal  Scheduled  35s   default-scheduler  Successfully assigned default/pod-antiaffinity to node2.noisedu.cn
  Normal  Pulled     30s   kubelet            Container image "10.0.0.55:80/mykubernetes/pod_test:v0.1" already present on machine
  Normal  Created    30s   kubelet            Created container pod-test
  Normal  Started    30s   kubelet            Started container pod-test
 
# 由于test权重低，所以调度到node2上面

　　

posted @   每天都在学习的自己  阅读(281)  评论(0)  编辑  收藏  举报
相关博文：
· Linux - K8S - 调度策略 - Node调度
· Linux - K8S - 调度策略 - 污点和容忍度
· K8S群集调度器
· Pod亲和与反亲和、污点与容忍、驱逐
· 关于K8S亲和性的解释
阅读排行：
· 阿里最新开源QwQ-32B，效果媲美deepseek-r1满血版，部署成本又又又降低了！
· 开源Multi-agent AI智能体框架aevatar.ai，欢迎大家贡献代码
· Manus重磅发布：全球首款通用AI代理技术深度解析与实战指南
· 被坑几百块钱后，我竟然真的恢复了删除的微信聊天记录！
· AI技术革命，工作效率10个最佳AI工具
点击右上角即可分享
微信分享提示