关于 Kubernetes中Pod的一些笔记(三)
【摘要】 钱比你想象的重要得多,超过20岁了就别整天活在梦里了,对于平凡的你来讲,钱就是你的尊严。
写在前面
- 学习
K8s
,刚把Pod
学完,整理笔记记忆 - 笔记主要是
Pod
的一些基本操作,偏实战,理论很少: - 笔记内容包括:
- 创建
Pod
的两种方式,相关镜像下载,重启机制 Pod
的详细信息,日志、命令运行等、生命周期等- 初始化
Pod
和静态Pod
Pod
的调度(选择器、指定节点、主机亲和性)- 节点的
coedon
与drain
- 节点的
taint
(污点)及容忍污点(tolerations
) - 部分地方使用了
ansible
,但是不影响阅读
- 创建
钱比你想象的重要得多,超过20岁了就别整天活在梦里了,对于平凡的你来讲,钱就是你的尊严。
六、Pod调度
调度的三个对象
待调度Pod列表:有多少个pod需要调度,即创建的pod列表
可用node列表:有那些节点可以参与调度,排除有污点,端口的一些node
调度算法
- 主机过滤
+ `NoDiskConflict`
+ `PodFitsResources`
+ `PodFitsPorts`
+ `MatchNodeSelector`
+ `HostName`
+ `NoVolumeZoneConflict`
+ `PodToleratesNodeTaints`
+ `CheckNodeMemoryPressure`
+ `CheckNodeDiskPressure`
+ `MaxEBSVolumeCount`
+ `MaxGCEPDVolumeCount`
+ `MaxAzureDiskVolumeCount`
+ `MatchInterPodAffinity`
+ `GeneralPredicates`
+ `NodeVolumeNodeConflic`
- 主机打分
分数项 | 公式 |
---|---|
LeastRequestedPriority | score=cpu ( ( capacity - sum ( requested ) ) * 10 / capacity) + memory ( ( capacity - sum ( requested) ) * 10 / capacity )/2 |
BalanceResourceAllocation | score = 10 -abs ( cpuFraction - memoryFraction ) * 10 |
CalculateSpreadPriority | Score = 10 * ((maxCount -counts)/ (maxCount)) |
手动指定pod的运行位置:
可以给node设置指定的标签,然后我们可以在创建pod里指定node标签
标签设置 | – |
---|---|
查看 | kubectl get nodes --show-labels |
设置 | kubectl label node node2 disktype=ssd |
取消 | kubectl label node node2 disktype |
所有节点设置 | kubectl label node all key=vale |
查看节点pod:kubectl get node --show-labels
给节点设置标签
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl label node vms82.liruilongs.github.io disktype=node1
node/vms82.liruilongs.github.io labeled
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl label node vms83.liruilongs.github.io disktype=node2
node/vms83.liruilongs.github.io labeled
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl get node --show-labels
NAME STATUS ROLES AGE VERSION LABELS
vms81.liruilongs.github.io Ready control-plane,master 45d v1.22.2 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,kubernetes.io/arch=amd64,kubernetes.io/hostname=vms81.liruilongs.github.io,kubernetes.io/os=linux,node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=,node.kubernetes.io/exclude-from-external-load-balancers=
vms82.liruilongs.github.io Ready <none> 45d v1.22.2 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=node1,kubernetes.io/arch=amd64,kubernetes.io/hostname=vms82.liruilongs.github.io,kubernetes.io/os=linux
vms83.liruilongs.github.io Ready <none> 45d v1.22.2 beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disktype=node2,kubernetes.io/arch=amd64,kubernetes.io/hostname=vms83.liruilongs.github.io,kubernetes.io/os=linux
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$
特殊的内置标签node-role.kubernetes.io/control-plane=,node-role.kubernetes.io/master=
,用于设置角色列roles
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl get node
NAME STATUS ROLES AGE VERSION
vms81.liruilongs.github.io Ready control-plane,master 45d v1.22.2
vms82.liruilongs.github.io Ready <none> 45d v1.22.2
vms83.liruilongs.github.io Ready <none> 45d v1.22.2
我们也可以做worker节点上设置
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl label nodes vms82.liruilongs.github.io node-role.kubernetes.io/worker1=
node/vms82.liruilongs.github.io labeled
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl label nodes vms83.liruilongs.github.io node-role.kubernetes.io/worker2=
node/vms83.liruilongs.github.io labeled
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$kubectl get node
NAME STATUS ROLES AGE VERSION
vms81.liruilongs.github.io Ready control-plane,master 45d v1.22.2
vms82.liruilongs.github.io Ready worker1 45d v1.22.2
vms83.liruilongs.github.io Ready worker2 45d v1.22.2
┌──[root@vms81.liruilongs.github.io]-[~/ansible]
└─$
选择器(nodeSelector
)方式
在特定节点上运行pod
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get nodes -l disktype=node2
NAME STATUS ROLES AGE VERSION
vms83.liruilongs.github.io Ready worker2 45d v1.22.2
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$vim pod-node2.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl apply -f pod-node2.yaml
pod/podnode2 created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
podnode2 1/1 Running 0 13m 10.244.70.60 vms83.liruilongs.github.io <none> <none>
pod-node2.yaml
apiVersion: v1
kind: Pod
metadata:
creationTimestamp: null
labels:
run: podnode2
name: podnode2
spec:
nodeSelector:
disktype: node2
containers:
- image: nginx
imagePullPolicy: IfNotPresent
name: podnode2
resources: {}
dnsPolicy: ClusterFirst
restartPolicy: Always
status: {}
指定节点名称(nodeName
)的方式
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$vim pod-node1.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl apply -f pod-node1.yaml
pod/podnode1 created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
podnode1 1/1 Running 0 36s 10.244.171.165 vms82.liruilongs.github.io <none> <none>
podnode2 1/1 Running 0 13m 10.244.70.60 vms83.liruilongs.github.io <none> <none>
pod-node1.yaml
apiVersion: v1
kind: Pod
metadata:
creationTimestamp: null
labels:
run: podnode1
name: podnode1
spec:
nodeName: vms82.liruilongs.github.io
containers:
- image: nginx
imagePullPolicy: IfNotPresent
name: podnode1
resources: {}
dnsPolicy: ClusterFirst
restartPolicy: Always
status: {}
当pod资源文件指定的节点标签,或者节点名不存在时,这个pod资源是无法创建成功的
主机亲和性
所谓主机亲和性,即在满足指定条件的节点上运行。分为硬策略(必须满足),软策略(最好满足)
硬策略(requiredDuringSchedulingIgnoredDuringExecution
)
pod-node-a.yaml
apiVersion: v1
kind: Pod
metadata:
creationTimestamp: null
labels:
run: podnodea
name: podnodea
spec:
containers:
- image: nginx
imagePullPolicy: IfNotPresent
name: podnodea
resources: {}
affinity:
nodeAffinity: #主机亲和性
requiredDuringSchedulingIgnoredDuringExecution: #硬策略
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- vms85.liruilongs.github.io
- vms84.liruilongs.github.io
dnsPolicy: ClusterFirst
restartPolicy: Always
status: {}
条件不满足,所以 Pending
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl apply -f pod-node-a.yaml
pod/podnodea created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get pods
NAME READY STATUS RESTARTS AGE
podnodea 0/1 Pending 0 8s
我梦修改一下
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$sed -i 's/vms84.liruilongs.github.io/vms83.liruilongs.github.io/' pod-node-a.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl apply -f pod-node-a.yaml
pod/podnodea created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
podnodea 1/1 Running 0 13s 10.244.70.61 vms83.liruilongs.github.io <none> <none>
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$
软策略(preferredDuringSchedulingIgnoredDuringExecution
)
pod-node-a-r.yaml
apiVersion: v1
kind: Pod
metadata:
creationTimestamp: null
labels:
run: podnodea
name: podnodea
spec:
containers:
- image: nginx
imagePullPolicy: IfNotPresent
name: podnodea
resources: {}
affinity:
nodeAffinity: #主机亲和性
preferredDuringSchedulingIgnoredDuringExecution: # 软策略
- weight: 2
preference:
matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- vms85.liruilongs.github.io
- vms84.liruilongs.github.io
dnsPolicy: ClusterFirst
restartPolicy: Always
status: {}
检查一下
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$vim pod-node-a-r.yaml
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl apply -f pod-node-a-r.yaml
pod/podnodea created
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl get pods -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
podnodea 1/1 Running 0 28s 10.244.70.62 vms83.liruilongs.github.io <none> <none>
运算符 | 描述 |
---|---|
In | 包含自, 比如上面的硬亲和就包含env_role=dev、env_role=test两种标签 |
NotIn | 和上面相反,凡是包含该标签的节点都不会匹配到 |
Exists | 存在里面和In比较类似,凡是有某个标签的机器都会被选择出来。使用Exists的operator的话,values里面就不能写东西了。 |
Gt | greater than的意思,表示凡是某个value大于设定的值的机器则会被选择出来。 |
Lt | less than的意思,表示凡是某个value小于设定的值的机器则会被选择出来。 |
DoesNotExists | 不存在该标签的节点 |
Annotations 的设置
Annotations 即注释,设置查看方式很简单
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl annotate nodes vms82.liruilongs.github.io "dest=这是一个工作节点"
node/vms82.liruilongs.github.io annotated
┌──[root@vms81.liruilongs.github.io]-[~/ansible/k8s-pod-create]
└─$kubectl describe nodes vms82.liruilongs.github.io
Name: vms82.liruilongs.github.io
Roles: worker1
Labels: beta.kubernetes.io/arch=amd64
beta.kubernetes.io/os=linux
disktype=node1
kubernetes.io/arch=amd64
kubernetes.io/hostname=vms82.liruilongs.github.io
kubernetes.io/os=linux
node-role.kubernetes.io/worker1=
Annotations: dest: 这是一个工作节点
kubeadm.alpha.kubernetes.io/cri-socket: /var/run/dockershim.sock
node.alpha.kubernetes.io/ttl: 0
projectcalico.org/IPv4Address: 192.168.26.82/24
projectcalico.org/IPv4IPIPTunnelAddr: 10.244.171.128
volumes.kubernetes.io/controller-managed-attach-detach: true
【版权声明】本文为华为云社区用户原创内容,转载时必须标注文章的来源(华为云社区)、文章链接、文章作者等基本信息, 否则作者和本社区有权追究责任。如果您发现本社区中有涉嫌抄袭的内容,欢迎发送邮件进行举报,并提供相关证据,一经查实,本社区将立刻删除涉嫌侵权内容,举报邮箱:
cloudbbs@huaweicloud.com
- 点赞
- 收藏
- 关注作者
评论(0)