k8s排程器
節點硬親和度：Pod排程時必須滿足規劃，如不滿足Pod狀態將是Penging狀態。
節點軟親和度：Pod排程時先按規則進行排程，如不滿足規則也會找一個不匹配的節點執行。
Pod硬親和度，軟親和度與節點硬親和度，軟親和度相似。
1.節點硬親和度(節點不滿足規則)
[root@k8s01 yaml]# cat pod-affinity01.yaml
apiVersion: v1
kind: Pod
metadata:
name: pod-01
spec:
affinity:
nodeAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
nodeSelectorTerms:
- matchExpressions:
- {key: zone,operator: In,values: ["one"]} --節點zone(鍵)標籤必須是one(值)
containers:
- name: pod-01
image: nginx:latest
imagePullPolicy: Never
[root@k8s01 yaml]# kubectl apply -f pod-affinity01.yaml
pod/pod-01 created
[root@k8s01 yaml]# kubectl get pods --show-labels
NAME READY STATUS RESTARTS AGE LABELS
pod-01 0/1 Pending 0 103s <none> --node節點不滿足pod標籤
[root@k8s01 yaml]# kubectl describe pods pod-01
。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。。
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling <unknown> default-scheduler 0/3 nodes are available: 3 node(s) didn't match node selector.
Warning FailedScheduling <unknown> default-scheduler 0/3 nodes are available: 3 node(s) didn't match node selector.
[root@k8s01 yaml]#
- matchExpressions:
- {key: zone,operator: In,values: ["one"]} --zone是鍵，operator：是表示式，one是值
- {key: zone,operator: In,values: ["one","two"]} --In 是包含
- {key: ssd,operator: Exists,values: []} --Exists是存在
2.節點硬親和度(節點滿足規則)
[root@k8s01 yaml]# kubectl label node k8s02 zone=one --在k8s02節點建立標籤
node/k8s02 labeled
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-01 1/1 Running 0 6m13s 10.244.1.37 k8s02 <none> <none> <none>
[root@k8s01 yaml]#
3.節點軟親和度
[root@k8s01 yaml]# cat pod-affinity02.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: pod-02
spec:
replicas: 3
selector:
matchLabels:
app: myapp
template:
metadata:
name: myapp-pod
labels:
app: myapp
spec:
affinity:
nodeAffinity:
preferredDuringSchedulingIgnoredDuringExecution:
- weight: 60 --60%的Pod會排程到zone標籤，one值這個節點上
preference:
matchExpressions:
- {key: zone,operator: In,values: ["one"]}
containers:
- name: pod-02
image: nginx:latest
imagePullPolicy: Never
[root@k8s01 yaml]# kubectl apply -f pod-affinity02.yaml
deployment.apps/pod-02 created
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide --滿足標籤可建立，不滿足也可建立
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-02-77d87986b-9bzg5 1/1 Running 0 16s 10.244.1.39 k8s02 <none> <none> app=myapp,pod-template-hash=77d87986b
pod-02-77d87986b-dckjq 1/1 Running 0 16s 10.244.2.42 k8s03 <none> <none> app=myapp,pod-template-hash=77d87986b
pod-02-77d87986b-z7v47 1/1 Running 0 16s 10.244.1.38 k8s02 <none> <none> app=myapp,pod-template-hash=77d87986b
[root@k8s01 yaml]#
4. Pod硬親和度(節點不滿足規則)
[root@k8s01 yaml]# kubectl get pods --show-labels
NAME READY STATUS RESTARTS AGE LABELS
nginx 1/1 Running 0 4s
app=web
[root@k8s01 yaml]# cat pod-affinity03.yaml
apiVersion: v1
kind: Pod
metadata:
name: pod-01
spec:
affinity:
podAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- {key: app,operator: In,values: ["web1"]} --標籤值不滿足
topologyKey: kubernetes.io/hostname
containers:
- name: pod-01
image: nginx:latest
imagePullPolicy: Never
[root@k8s01 yaml]# kubectl apply -f pod-affinity03.yaml
pod/pod-01 created
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
nginx 1/1 Running 0 8m9s 10.244.1.42 k8s02 <none> <none> app=web
pod-01 0/1
Pending 0 28s <none> <none> <none> <none> <none>
[root@k8s01 yaml]#
5.Pod硬親和度 (節點滿足規則)
[root@k8s01 yaml]# kubectl get pods --show-labels --檢視pod標籤
NAME READY STATUS RESTARTS AGE LABELS
nginx 1/1 Running 0 4s
app=web
[root@k8s01 yaml]# cat pod-affinity04.yaml
apiVersion: v1
kind: Pod
metadata:
name: pod-01
spec:
affinity:
podAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- {key: app,operator: In,values: ["web"]}
topologyKey: kubernetes.io/hostname
containers:
- name: pod-01
image: nginx:latest
imagePullPolicy: Never
[root@k8s01 yaml]# kubectl apply -f pod-affinity04.yaml
pod/pod-01 created
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide --建立的新Pod會找app=web的標籤
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
nginx 1/1 Running 0 4m14s 10.244.1.42
k8s02 <none> <none> app=web
pod-01 1/1
Running 0 17s 10.244.1.43
k8s02 <none> <none> <none>
[root@k8s01 yaml]#
6.Pod軟親和度
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
nginx 1/1 Running 0 14m 10.244.1.42 k8s02 <none> <none> app=web
[root@k8s01 yaml]# cat pod-affinity04.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: pod-02
spec:
replicas: 3
selector:
matchLabels:
app: myapp
template:
metadata:
name: myapp-pod
labels:
app: myapp
spec:
affinity:
podAffinity:
preferredDuringSchedulingIgnoredDuringExecution:
- weight: 60
podAffinityTerm:
labelSelector:
matchExpressions:
- {key: app,operator: In,values: ["web12"]}
topologyKey: zone
containers:
- name: pod-02
image: nginx:latest
imagePullPolicy: Never
[root@k8s01 yaml]# kubectl get pods --show-labels -o wide --不滿足Pod軟親和度也可建立
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
nginx 1/1 Running 0 16m 10.244.1.42 k8s02 <none> <none> app=web
pod-02-6f96fbfdf-52gbf 1/1 Running 0 11s 10.244.2.43
k8s03 <none> <none> app=myapp,pod-template-hash=6f96fbfdf
pod-02-6f96fbfdf-dl5z5 1/1 Running 0 11s 10.244.1.44
k8s02 <none> <none> app=myapp,pod-template-hash=6f96fbfdf
pod-02-6f96fbfdf-f8bzn 1/1 Running 0 11s 10.244.0.55
k8s01 <none> <none> app=myapp,pod-template-hash=6f96fbfdf
[root@k8s01 yaml]#
7.定義汙點和容忍度
NoSchedule：不能容忍此汙點的新Pod對像，不能排程到當前節點，強制性。
PreferNoSchedule：新Pod儘量不排程到此節點，若無其它節點可以排程，也允許使用當前節點排程，柔和性。
NoExecute：新Pod不能排程到當前節點，如果當前節點存在Pod也會被驅逐，強制性。
[root@k8s01 yaml]# kubectl describe node k8s02 | grep -i taints --檢視汙點
Taints: <none>
[root@k8s01 yaml]# kubectl taint node k8s02 node-type=production:NoSchedule --為k8s02節點打汙點
node/k8s02 tainted
[root@k8s01 yaml]# kubectl describe node k8s02 | grep -i taints
Taints: node-type=production:NoSchedule
[root@k8s01 yaml]# cat pod-affinity05.yaml
apiVersion: apps/v1 kind: Deployment metadata: name: pod-02 spec: replicas: 3 selector: matchLabels: app: myapp template: metadata: name: myapp-pod labels: app: myapp spec: containers: - name: pod-02 image: nginx:latest imagePullPolicy: Never
[root@k8s01 yaml]# kubectl apply -f pod-affinity05.yaml
deployment.apps/pod-02 created
[root@k8s01 yaml]# kubectl get pods -o wide --show-labels --k8s02節點已打汙點，所以不能排程Pod對像
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-02-5c54dc6489-j7tsh 1/1 Running 0 29s 10.244.0.56
k8s01 <none> <none> app=myapp,pod-template-hash=5c54dc6489
pod-02-5c54dc6489-sk5dm 1/1 Running 0 29s 10.244.2.44
k8s03 <none> <none> app=myapp,pod-template-hash=5c54dc6489
pod-02-5c54dc6489-sn5wd 1/1 Running 0 29s 10.244.2.45
k8s03 <none> <none> app=myapp,pod-template-hash=5c54dc6489
[root@k8s01 yaml]
8.刪除汙點
[root@k8s01 yaml]# kubectl taint node k8s02 node-type:NoSchedule- --刪除 NoSchedule汙點
node/k8s02 untainted
[root@k8s01 yaml]# kubectl taint node k8s02 node-type=production:PreferNoSchedule --打上汙點
node/k8s02 tainted
[root@k8s01 yaml]# kubectl describe node k8s02 | grep -i node-type
Taints: node-type=production:PreferNoSchedule
[root@k8s01 yaml]# kubectl taint node k8s02 node-type:PreferNoSchedule- --刪除
PreferNoSchedule汙點
node/k8s02 untainted
[root@k8s01 yaml]# kubectl describe node k8s02 | grep -i node-type
[root@k8s01 yaml]# kubectl taint node k8s02 node-type- --刪除所有汙點
node/k8s02 untainted
[root@k8s01 yaml]# kubectl describe node k8s02 | grep -i node-type
[root@k8s01 yaml]#
9. 驅逐節點上的Pod
[root@k8s01 yaml]# kubectl cordon k8s03 --將新建立的pod不能排程到k8s03節點，以前的Pod不受影響。
node/k8s03 cordoned
[root@k8s01 yaml]# kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s01 Ready master 71d v1.16.0
k8s02 Ready <none> 70d v1.16.0
k8s03 Ready,
SchedulingDisabled <none> 30d v1.16.0
[root@k8s01 yaml]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod-02-5c54dc6489-442kx 1/1 Running 0 12m 10.244.2.48
k8s03 <none> <none>
pod-02-5c54dc6489-92l8m 1/1 Running 0 12m 10.244.2.49
k8s03 <none> <none>
pod-02-5c54dc6489-k4bc7 1/1 Running 0 12m 10.244.0.58 k8s01 <none> <none>
[root@k8s01 yaml]# kubectl uncordon k8s03 --釋放排程規則
node/k8s03 uncordoned
[root@k8s01 yaml]# kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s01 Ready master 71d v1.16.0
k8s02 Ready <none> 70d v1.16.0
k8s03 Ready <none> 30d v1.16.0
[root@k8s01 yaml]# kubectl drain k8s03 --ignore-daemonsets --去除k8s03節點上的Pod
node/k8s03 already cordoned
WARNING: ignoring DaemonSet-managed Pods: kube-system/kube-flannel-ds-amd64-cg795, kube-system/kube-proxy-h4dkf
evicting pod "pod-02-5c54dc6489-92l8m"
evicting pod "pod-02-5c54dc6489-442kx"
pod/pod-02-5c54dc6489-92l8m evicted
pod/pod-02-5c54dc6489-442kx evicted
node/k8s03 evicted
[root@k8s01 yaml]# kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s01 Ready master 71d v1.16.0
k8s02 Ready <none> 70d v1.16.0
k8s03 Ready,
SchedulingDisabled <none> 30d v1.16.0
[root@k8s01 yaml]# kubectl get pods -o wide --以前k8s03節點的Pod全部轉移到其它節點上
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pod-02-5c54dc6489-k4bc7 1/1 Running 0 14m 10.244.0.58
k8s01 <none> <none>
pod-02-5c54dc6489-mxk46 1/1 Running 0 25s 10.244.1.46
k8s02 <none> <none>
pod-02-5c54dc6489-vmb8l 1/1 Running 0 25s 10.244.1.45
k8s02 <none> <none>
[root@k8s01 yaml]# kubectl uncordon k8s03 --恢復排程
node/k8s03 uncordoned
[root@k8s01 yaml]# kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s01 Ready master 71d v1.16.0
k8s02 Ready <none> 70d v1.16.0
k8s03 Ready <none> 30d v1.16.0
[root@k8s01 yaml]#
來自 “ ITPUB部落格 ” ，連結：http://blog.itpub.net/25854343/viewspace-2664507/，如需轉載，請註明出處，否則將追究法律責任。
相關文章
- k8s排程器介紹（排程框架版本）K8S框架
- k8s 排程 GPUK8SGPU
- Flink排程之排程器、排程策略、排程模式模式
- k8s之pod排程K8S
- K8S 高階排程方式K8S
- Go排程器系列（2）巨集觀看排程器Go
- 從零開始入門 K8s | 排程器的排程流程和演算法介紹K8S演算法
- Go語言排程器之主動排程(20)Go
- Go排程器系列（3）圖解排程原理Go圖解
- goroutine 排程器（scheduler）Go
- MYSQL事件排程器MySql事件
- Go runtime 排程器精講（二）：排程器初始化Go
- 排程器簡介，以及Linux的排程策略Linux
- Go語言排程器之排程main goroutine（14）GoAI
- Pod的排程是由排程器（kube-scheduler）
- Go runtime 排程器精講（五）：排程策略Go
- 也談goroutine排程器Go
- 解決k8s排程不均衡問題K8S
- [典藏版] Golang 排程器 GMP 原理與排程全分析Golang
- K8S原來如此簡單（六）Pod排程K8S
- K8s 容器的定向排程與親和性K8S
- Linux I/O排程器Linux
- RxJava排程器的選擇RxJava
- Go Runtime 的排程器Go
- 【MySQL】事件排程器 (Event Scheduler)MySql事件
- 使用 K8s 進行作業排程實戰分享K8S
- 在K8S中，Pod常見排程方式有哪些?K8S
- MySQL中的事件排程器EVENTMySql事件
- [譯] RxJS 排程器入門JS
- Hadoop排程器原理解析Hadoop
- K8s中的排程程式是如何工作的？ - danieleK8S
- Linux核心排程分析（程式排程）Linux
- k8s中檢視預設排程器開啟的外掛以及權重的含義K8S
- Kubernetes 排程器實現初探
- LeetCode 621 任務排程器LeetCode
- 621. 任務排程器 （構造）
- 力扣 621. 任務排程器力扣
- golang 原始碼分析之scheduler排程器Golang原始碼