nodeSelector
作用及说明
参考文档： https://kubernetes.io/zh-cn/docs/concepts/scheduling-eviction/assign-pod-node/
#nodeSelector 是节点选择约束的最简单推荐形式。你可以将 nodeSelector 字段添加到 Pod 的规约中设置你希望目标节点所具有的节点标签。 Kubernetes 只会将 Pod 调度到拥有你所指定的每个标签的节点上。简单来说 nodeSelector 是通过匹配节点标签，让 Pod 只能调度到指定类型节点上的最简单调度方式。
#使用说明
[root@k8s-master01 ~]# kubectl explain deploy.spec.template.spec.nodeSelector
GROUP: apps
KIND: Deployment
VERSION: v1
FIELD: nodeSelector <map[string]string>
DESCRIPTION:
NodeSelector is a selector which must be true for the pod to fit on a node.
Selector which must match a node's labels for the pod to be scheduled on
that node. More info:
https://kubernetes.io/docs/concepts/configuration/assign-pod-node/
nodeselector实战案例
# 查看node节点的标签
[root@k8s-master01 ~]# kubectl get nodes --show-labels
NAME STATUS ROLES AGE VERSION LABELS
k8s-master01.dinginx.org Ready control-plane 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master01.dinginx.org,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node.kubernetes.io/exclude-from-external-load-balancers=
k8s-node01.dinginx.org Ready <none> 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.dinginx.org,kubernetes.io/os=linux
k8s-node02.dinginx.org Ready <none> 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02.dinginx.org,kubernetes.io/os=linux
# 为node02节点添加新标签
[root@k8s-master01 ~]# kubectl label node k8s-node01.dinginx.org disktype=ssd
[root@k8s-master01 ~]# kubectl label node k8s-node02.dinginx.org disktype=ssd
[root@k8s-master01 ~]# kubectl get nodes --show-labels
NAME STATUS ROLES AGE VERSION LABELS
k8s-master01.dinginx.org Ready control-plane 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-master01.dinginx.org,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node.kubernetes.io/exclude-from-external-load-balancers=
k8s-node01.dinginx.org Ready <none> 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.dinginx.org,kubernetes.io/os=linux
k8s-node02.dinginx.org Ready <none> 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=ssd,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node02.dinginx.org,kubernetes.io/os=linux
#***********
查看node节点是否存在污点，如存在请先删除污点或增加污点容忍
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl describe nodes |grep Taints
Taints: app=dinginx001:NoSchedule
Taints: <none>
Taints: <none>
#***********
# 编写资源清单
[root@k8s-master01 /data/manifests/pod-scheduler]# cat 07-scheduler-nodeSelector.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: dinginx-deploy-nodeselector
spec:
replicas: 3
selector:
matchLabels:
app: dinginx-deploy-nodeselector
template:
metadata:
labels:
app: dinginx-deploy-nodeselector
spec:
containers:
- image: harbor.dinginx.org/dinginx/nginx:latest
name: nginx
imagePullPolicy: IfNotPresent
# 创建标签选择器，表明该 Pod 将被调度到有 disktype=ssd 标签的节点。
nodeSelector:
disktype: ssd
`#查看资源信息`，资源被调度到k8s-node02.dinginx.org节点上
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
dinginx-deploy-nodeselector-7f5c76ff85-7bnsq 1/1 Running 0 4m9s 10.244.2.235 k8s-node02.dinginx.org <none> <none> app=dinginx-deploy-nodeselector,pod-template-hash=7f5c76ff85
dinginx-deploy-nodeselector-7f5c76ff85-fvzw7 1/1 Running 0 4m9s 10.244.2.236 k8s-node02.dinginx.org <none> <none> app=dinginx-deploy-nodeselector,pod-template-hash=7f5c76ff85
dinginx-deploy-nodeselector-7f5c76ff85-hcp9n 1/1 Running 0 4m9s 10.244.2.238 k8s-node02.dinginx.org <none> <none> app=dinginx-deploy-nodeselector,pod-template-hash=7f5c76ff85
affinity（亲和性）
作用及说明
参考链接：https://kubernetes.io/zh-cn/docs/concepts/scheduling-eviction/assign-pod-node/`#affinity`-and-anti-affinity
#使用说明
[root@k8s-master01 ~]# kubectl explain po.spec.affinity
KIND: Pod
VERSION: v1
FIELD: affinity <Affinity>
DESCRIPTION:
If specified, the pod's scheduling constraints
Affinity is a group of affinity scheduling rules.
FIELDS:
nodeAffinity <NodeAffinity>
Describes node affinity scheduling rules for the pod.
podAffinity <PodAffinity>
Describes pod affinity scheduling rules (e.g. co-locate this pod in the same
node, zone, etc. as some other pod(s)).
podAntiAffinity <PodAntiAffinity>
Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in
the same node, zone, etc. as some other pod(s)).
nodeAffinity（节点亲和性）案例
作用及说明
# 基于节点的标签进行调度，相比较nodeselector可实现多值匹配。
`## 使用说明`
[root@k8s-master01 ~]# kubectl explain po.spec.affinity.nodeAffinity
KIND: Pod
VERSION: v1
FIELD: nodeAffinity <NodeAffinity>
DESCRIPTION:
Describes node affinity scheduling rules for the pod.
Node affinity is a group of node affinity scheduling rules.
FIELDS:
`#软限制`
preferredDuringSchedulingIgnoredDuringExecution <[]PreferredSchedulingTerm>
The scheduler will prefer to schedule pods to nodes that satisfy the
affinity expressions specified by this field, but it may choose a node that
violates one or more of the expressions. The node that is most preferred is
the one with the greatest sum of weights, i.e. for each node that meets all
of the scheduling requirements (resource request, requiredDuringScheduling
affinity expressions, etc.), compute a sum by iterating through the elements
of this field and adding "weight" to the sum if the node matches the
corresponding matchExpressions; the node(s) with the highest sum are the
most preferred.
`#强制限制`，推荐使用
requiredDuringSchedulingIgnoredDuringExecution <NodeSelector>
If the affinity requirements specified by this field are not met at
scheduling time, the pod will not be scheduled onto the node. If the
affinity requirements specified by this field cease to be met at some point
during pod execution (e.g. due to an update), the system may or may not try
to eventually evict the pod from its node.
`#配置`
[root@k8s-master01 ~]# kubectl explain po.spec.affinity.nodeAffinity.requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms.matchExpressions
requiredDuringSchedulingIgnoredDuringExecution(硬限制)实战案例
`#资源清单`，实现资源调度到master及node02节点，master节点需配置容忍度
[root@k8s-master01 /data/manifests/pod-scheduler]# cat 08-scheduler-nodeAffinity.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: dinginx-deploy-nodeaffinity
spec:
replicas: 3
selector:
matchLabels:
app: dinginx-deploy-nodeaffinity
template:
metadata:
labels:
app: dinginx-deploy-nodeaffinity
spec:
# 配置容忍度
tolerations:
# 容忍master污点规则
- key: app
value: dinginx001
effect: NoSchedule
containers:
- name: nginx
image: harbor.dinginx.org/dinginx/nginx:latest
imagePullPolicy: IfNotPresent
# 亲和性调度配置
affinity:
# 节点亲和性（控制Pod调度到哪些节点）
nodeAffinity:
# 硬性规则（必须满足）
# Pod调度时必须满足，否则不调度
requiredDuringSchedulingIgnoredDuringExecution:
# 注意：这里必须是数组
nodeSelectorTerms:
# 标签匹配表达式（可以写多个）
- matchExpressions:
# 条件1
# 节点标签 key
- key: disktype
# 操作符（In表示包含）
operator: In
values:
- ssd
# 相比较nodeselector可实现多值匹配
- dinginx
`#验证资源`，资源被成功调度到master和node02节点
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl apply -f 08-scheduler-nodeAffinity.yaml
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-nodeaffinity-7668b4556d-bdlz2 1/1 Running 0 10s 10.244.0.30 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7668b4556d-hcl82 1/1 Running 0 7s 10.244.0.31 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7668b4556d-j4p6s 1/1 Running 0 8s 10.244.2.165 k8s-node02.dinginx.org <none> <none>
# 匹配多值后验证，这里需提前给node02打标签disktype=dinginx
[root@k8s-master01 ~]# kubectl get nodes k8s-node01.dinginx.org --show-labels |grep disktype=dinginx
k8s-node01.dinginx.org Ready <none> 135d v1.35.0 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=dinginx,k8s.kuboard.cn/role=etcd,kubernetes.io/arch=amd64,kubernetes.io/hostname=k8s-node01.dinginx.org,kubernetes.io/os=linux
`#验证资源`
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-nodeaffinity-d6776db5-cl7bk 1/1 Running 0 7s 10.244.0.32 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-d6776db5-gz2mn 1/1 Running 0 6s 10.244.2.171 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-d6776db5-qlc44 1/1 Running 0 4s 10.244.1.162 k8s-node01.dinginx.org <none> <none>
preferredDuringSchedulingIgnoredDuringExecution(软限制)实战案例
`#使用说明`，
[root@k8s-master01 ~]# kubectl explain po.spec.affinity.podAffinity.preferredDuringSchedulingIgnoredDuringExecution
KIND: Pod
VERSION: v1
FIELD: preferredDuringSchedulingIgnoredDuringExecution <[]WeightedPodAffinityTerm>
DESCRIPTION:
The scheduler will prefer to schedule pods to nodes that satisfy the
affinity expressions specified by this field, but it may choose a node that
violates one or more of the expressions. The node that is most preferred is
the one with the greatest sum of weights, i.e. for each node that meets all
of the scheduling requirements (resource request, requiredDuringScheduling
affinity expressions, etc.), compute a sum by iterating through the elements
of this field and adding "weight" to the sum if the node has pods which
matches the corresponding podAffinityTerm; the node(s) with the highest sum
are the most preferred.
The weights of all of the matched WeightedPodAffinityTerm fields are added
per-node to find the most preferred node(s)
FIELDS:
podAffinityTerm <PodAffinityTerm> -required-
Required. A pod affinity term, associated with the corresponding weight.
weight <integer> -required-
weight associated with matching the corresponding podAffinityTerm, in the
range 1-100.
# 删除所有的标签
[root@k8s-master01 ~]# kubectl label node --all disktype-
# 创建资源清单
[root@k8s-master01 /data/manifests/pod-scheduler]# cat 08-scheduler-nodeAffinity.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: dinginx-deploy-nodeaffinity
spec:
replicas: 10
selector:
matchLabels:
app: dinginx-deploy-nodeaffinity
template:
metadata:
labels:
app: dinginx-deploy-nodeaffinity
spec:
tolerations:
- key: app
value: dinginx001
effect: NoSchedule
containers:
- name: nginx
image: harbor.dinginx.org/dinginx/nginx:latest
imagePullPolicy: IfNotPresent
# 亲和性调度配置
affinity:
# 节点亲和性（控制Pod调度到哪些节点）
nodeAffinity:
# 软限制规则（根据权重）
preferredDuringSchedulingIgnoredDuringExecution:
# 定义优先权策略，与相应权重关联的节点选择器项。
- preference:
matchExpressions:
- key: app
values:
- "ssd"
- "dinginx"
operator: In
# 设置权重，有效值为: 1-100
weight: 50
- preference:
matchExpressions:
- key: app
values:
- "dinginx001"
operator: In
weight: 80
`#打标签`
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl label nodes k8s-node01.dinginx.org app=ssd
node/k8s-node01.dinginx.org labeled
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl label nodes k8s-node02.dinginx.org app=dinginx001
node/k8s-node02.dinginx.org labeled
`#验证数据`，资源尽可能的网node02调度
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide |sort -t6
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-nodeaffinity-7b97dbdfc6-6bsxb 1/1 Running 0 2m6s 10.244.2.192 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-7mvmq 1/1 Running 0 2m6s 10.244.2.195 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-8dl4p 1/1 Running 0 2m6s 10.244.0.37 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-8dqp4 1/1 Running 0 2m6s 10.244.1.169 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-d978v 1/1 Running 0 2m6s 10.244.2.193 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-kvs94 1/1 Running 0 2m6s 10.244.0.39 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-s8nhd 1/1 Running 0 2m6s 10.244.0.38 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-s8wmd 1/1 Running 0 2m6s 10.244.1.170 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-v4rmw 1/1 Running 0 2m6s 10.244.2.194 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-nodeaffinity-7b97dbdfc6-zt226 1/1 Running 0 2m6s 10.244.0.40 k8s-master01.dinginx.org <none> <none>
一句话也就是说 硬限制：不满足就绝对不调度；软限制：尽量满足，不满足也可以调度。
podAffinity（Pod的亲和性）
软限制和硬限制参考nodeAffinity的相关配置：略
# 使用说明
[root@k8s-master01 ~]# kubectl explain po.spec.affinity.podAffinity
`#...`
FIELDS:
`#软限制`
preferredDuringSchedulingIgnoredDuringExecution <[]WeightedPodAffinityTerm>
`#硬性限制`
requiredDuringSchedulingIgnoredDuringExecution <[]PodAffinityTerm>
`#打标签`
[root@k8s-master01 ~]# kubectl label nodes k8s-master01.dinginx.org dc=beijing
[root@k8s-master01 ~]# kubectl label nodes k8s-node01.dinginx.org dc=shanghai
[root@k8s-master01 ~]# kubectl label nodes k8s-node02.dinginx.org dc=guangzhou
# 准备测试资源
[root@k8s-master01 /data/manifests/pod-scheduler]# cat test.yaml
apiVersion: v1
kind: Pod
metadata:
labels:
app: dinginx-deploy-podaffinity
name: dinginx-deploy-podaffinity
spec:
nodeName: k8s-node01.dinginx.org
containers:
- image: nginx
name: dinginx-deploy-podaffinity
dnsPolicy: ClusterFirst
`## 下面pod亲和性会调度到node01节点`
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get -f test.yaml -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-podaffinity 1/1 Running 0 3m31s 10.244.1.192 k8s-node01.dinginx.org <none> <none>
`#配置pod的亲和性`，创建资源清单
[root@k8s-master01 /data/manifests/pod-scheduler]# cat 09-scheduler-podAffinity.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: dinginx-deploy-podaffinity
spec:
replicas: 5
selector:
matchLabels:
app: dinginx-deploy-podaffinity
template:
metadata:
labels:
app: dinginx-deploy-podaffinity `#和测试pod标签一致`
spec:
tolerations:
- key: app
value: dinginx001
effect: NoSchedule
containers:
- name: nginx
image: harbor.dinginx.org/dinginx/nginx:latest
imagePullPolicy: IfNotPresent
# 亲和性调度配置
affinity:
# 节点亲和性（控制Pod调度到哪些节点）
podAffinity:
`#配置硬性限制`，不匹配不调度
requiredDuringSchedulingIgnoredDuringExecution:
# 指定拓扑域
- topologyKey: dc
# 基于pod关联
labelSelector:
# 基于标签匹配pod
matchLabels:
app: dinginx-deploy-podaffinity
# 创建并验证资源，pod全部和测试pod一致调度到node01节点
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl apply -f 09-scheduler-podAffinity.yaml
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-podaffinity 1/1 Running 0 5m18s 10.244.1.192 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podaffinity-5db7878587-5fsjb 1/1 Running 0 4m48s 10.244.1.195 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podaffinity-5db7878587-jf59n 1/1 Running 0 4m47s 10.244.1.194 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podaffinity-5db7878587-nqdjc 1/1 Running 0 4m48s 10.244.1.196 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podaffinity-5db7878587-s27x8 1/1 Running 0 4m48s 10.244.1.193 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podaffinity-5db7878587-x24t9 1/1 Running 0 4m47s 10.244.1.197 k8s-node01.dinginx.org <none> <none>
podAntiAffinity（Pod的反亲和性）
软限制和硬限制参考nodeAffinity的相关配置：略
`#使用说明`
[root@k8s-master01 ~]# kubectl explain po.spec.affinity.podAntiAffinity
KIND: Pod
VERSION: v1
FIELD: podAntiAffinity <PodAntiAffinity>
DESCRIPTION:
Describes pod anti-affinity scheduling rules (e.g. avoid putting this pod in
the same node, zone, etc. as some other pod(s)).
Pod anti affinity is a group of inter pod anti affinity scheduling rules.
FIELDS:
`#软限制`
preferredDuringSchedulingIgnoredDuringExecution <[]WeightedPodAffinityTerm>
`#硬性限制`
requiredDuringSchedulingIgnoredDuringExecution <[]PodAffinityTerm>
`#资源清单`
[root@k8s-master01 /data/manifests/pod-scheduler]# cat 10-scheduler-podAntiAffinity.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: dinginx-deploy-podantiaffinity
spec:
replicas: 5
selector:
matchLabels:
app: dinginx-deploy-podantiaffinity
template:
metadata:
labels:
app: dinginx-deploy-podantiaffinity
spec:
tolerations:
- key: app
value: dinginx001
effect: NoSchedule
containers:
- name: nginx
image: harbor.dinginx.org/dinginx/nginx:latest
imagePullPolicy: IfNotPresent
# 亲和性调度配置
affinity:
# 节点反亲和性（控制Pod调度到哪些节点）
podAntiAffinity:
`#配置硬性限制`，不匹配不调度
requiredDuringSchedulingIgnoredDuringExecution:
# 指定拓扑域
- topologyKey: dc
# 基于pod关联
labelSelector:
# 基于标签表达式匹配pod
matchExpressions:
- key: app
values: ["dinginx-deploy-podantiaffinity"]
operator: In
# 验证资源，发现两个pod为pending状态，是因为测试环境有三个节点，配置强制反亲和性后每个节点只能调度一个pod，所以超出部分无法调度
[root@k8s-master01 /data/manifests/pod-scheduler]# kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
dinginx-deploy-podantiaffinity-8558dcf684-4gj64 1/1 Running 0 10s 10.244.0.51 k8s-master01.dinginx.org <none> <none>
dinginx-deploy-podantiaffinity-8558dcf684-g76dz 1/1 Running 0 10s 10.244.1.200 k8s-node01.dinginx.org <none> <none>
dinginx-deploy-podantiaffinity-8558dcf684-l6mzq 0/1 Pending 0 10s <none> <none> <none> <none>
dinginx-deploy-podantiaffinity-8558dcf684-p8gxg 1/1 Running 0 10s 10.244.2.230 k8s-node02.dinginx.org <none> <none>
dinginx-deploy-podantiaffinity-8558dcf684-rdr2k 0/1 Pending 0 10s <none> <none> <none> <none>
通过以上实战案例，我们详细演示了如何从最简单的 nodeSelector 过渡到功能更强大的 Affinity 机制。在Kubernetes调度实践中，节点亲和性允许你根据节点标签进行更灵活（硬限制或软偏好）的调度控制，而Pod亲和性与反亲和性则能让你基于其他Pod的分布来安排新Pod，这对于实现高可用部署、服务隔离等场景至关重要。掌握这些机制能极大提升集群资源管理的稳定性和效率。如果你在配置过程中遇到问题，欢迎来云栈社区与大家一起交流探讨。
发表于 2026-2-11 10:56:35
|
查看: 32|
回复: 0
