Kubernetes学习目录
1、基础知识
1.1、什么是Pod调度
所谓的pod调度，主要说的是pod彼此之间的亲和性，也就是说，哪些pod应该在一起。
比如：我们的k8s集群的节点分布在不同的区域或者不同的机房
当服务A和服务B需要高效的交流数据的话，要求部署在同一个区域或者同一机房的时候。
当服务A需要做冗余操作，那么多个服务A必须在不同的位置
1.2、属性解析
kubectl explain pod.spec.affinity.podAffinity
requiredDuringSchedulingIgnoredDuringExecution -- 硬亲和性：
labelSelector 选择跟那组Pod亲和，前提得知道如何判断
namespaces 选择哪个命名空间进行条件匹配
topologyKey 指定节点上的哪个键，这是一个必选项
注意：这三个条件是一个逻辑与的关系
preferredDuringSchedulingIgnoredDuringExecution -- 软亲和性：
podAffinityTerm # 与权重关联的亲和选项，这是一个必选项
labelSelector
namespaces # 表示仅在指定的命名空间中查找
topologyKey # 先查询有该标签值，再去指定节点上调度对应的标签关键字
weight # 权重，这是一个必选项
2、Pod调度-affinity.podAffinity-实践
2.1、硬亲和性-requiredDuringSchedulingIgnoredDuringExecution
2.1.1、需求
创建两个在不同主机上的pod，然后亲和pod和标签env=test的pod部署在一起
2.1.2、给node节点创建标签
kubectl label node node2 env=test
kubectl label node node1 env=dev
2.1.3、定义资源配置清单且应用【使用节点标签选择器在每个节点创建一个Pod】
kubectl apply -f - <<EOF
apiVersion: v1
kind: Pod
metadata:
name: pod-dev
labels:
env: dev
spec:
containers:
- name: pod-test
image: 192.168.10.33:80/k8s/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: 192.168.10.33:80/k8s/pod_test:v0.1
imagePullPolicy: IfNotPresent
nodeSelector:
env: test
EOF
# 该pod的调度策略，env: test调度到node2，env: dev调度到node1
2.1.4、查询pod运行情况
master1 ~]# kubectl get pod -o wide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-dev 1/1 Running 0 38m 10.244.3.51 node1 <none> <none> env=dev
pod-test 1/1 Running 0 38m 10.244.4.84 node2 <none> <none> env=test
2.1.5、定义资源配置清单并且应用【创建一个pod，使用节点硬亲和性调度】
kubectl apply -f - <<EOF
apiVersion: v1
kind: Pod
metadata:
name: pod-affinity
labels:
env: test
spec:
containers:
- name: pod-test
image: 192.168.10.33:80/k8s/pod_test:v0.1
imagePullPolicy: IfNotPresent
affinity:
podAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- {key: env, operator: In, values: ["test"]}
namespaces: ["default"]
topologyKey: kubernetes.io/hostname
EOF
2.1.6、属性解析
说明解析：
1、给这个pod定义我们调度的标签，方便我们查询时，一眼能看出与哪个pod调度在一起
2、使用pod硬亲和性，表达式匹配标签，拓扑结构使用主机名类型
属性详解：
topologyKey的来源：
]# kubectl get nodes -o yaml | grep hostname
kubernetes.io/hostname: master1
kubernetes.io/hostname: master2
kubernetes.io/hostname: master3
kubernetes.io/hostname: node1
kubernetes.io/hostname: node2
其实指定 kubernetes.io/hostname 和 主机名 效果是一样的，只不过一个是自动获取，一个是手工指定
2.1.7、查询调度的结果
]# 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 8m39s 10.244.4.85 node2 <none> <none> env=test
pod-dev 1/1 Running 0 53m 10.244.3.51 node1 <none> <none> env=dev
pod-test 1/1 Running 0 53m 10.244.4.84 node2 <none> <none> env=test
# 查询发现，调度一致，都是把env=test调度到node2节点上
2.2、软亲和性-preferredDuringSchedulingIgnoredDuringExecution
2.2.1、需求
软亲和性类型，使用pod调度权重方式选择合适的pod调度到相同的节点。
2.2.2、给node节点创建标签
kubectl label node node2 env=test
kubectl label node node1 env=dev
2.2.3、定义资源配置清单
kubectl apply -f - <<EOF
apiVersion: v1
kind: Pod
metadata:
name: pod-affinity
spec:
containers:
- name: pod-test
image: 192.168.10.33:80/k8s/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
EOF
解析：
1、使用pod软亲和性调度，按权重选择最优符合的pod调度相同的节点。
2、拓朴类型：使用主机名
2.2.4、查询调度的结果
]# 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 9s 10.244.3.52 node1 <none> <none> <none>
pod-dev 1/1 Running 0 107m 10.244.3.51 node1 <none> <none> env=dev
pod-test 1/1 Running 0 107m 10.244.4.84 node2 <none> <none> env=test
分析：
被调度到权重最高的env=dev的pod相同的节点，说明标签权重pod调度是生效的。
2.2.5、删除env=dev的pod，重新创建pod-affinity对象，观察是否被调度到标签权重小pod相同的节点
# 删除pod
kubectl delete pod pod-dev
kubectl delete pod pod-affinity
重新创建pod,请重新执行一下：2.2.3、定义资源配置清单
master1 ~]# 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 10s 10.244.4.86 node2 <none> <none> <none>
pod-test 1/1 Running 0 114m 10.244.4.84 node2 <none> <none> env=test
# 可以看到，被调度到权重最小的pod节点一起
3、redis与pod软亲和性、硬亲和性-实践
3.1、redis硬亲和性-实践
3.1.1、需求
Pod必须把数据存储到redis中，必须在一起，提高写入速度。
3.1.2、定义资源配置清单且应用【创建redis的pod】
kubectl apply -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
name: redis
spec:
replicas: 1
selector:
matchLabels:
app: redis
template:
metadata:
labels:
app: redis
spec:
containers:
- name: redis
image: 192.168.10.33:80/k8s/redis:latest
imagePullPolicy: IfNotPresent
EOF
3.1.3、查询redis创建的结果
master1 ~]# kubectl get pod -o wide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
redis-56f97b7f57-x82dj 1/1 Running 0 10s 10.244.3.53 node1 <none> <none> app=redis,pod-template-hash=56f97b7f57
3.1.4、定义资源配置清单【创建测试pod，使用标签调度到跟pod标签匹配的相同节点上】
kubectl apply -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
name: pod-affinity-required
spec:
replicas: 4
selector:
matchLabels:
app: pod-test
template:
metadata:
labels:
app: pod-test
spec:
containers:
- name: pod-test
image: 192.168.10.33:80/k8s/pod_test:v0.1
imagePullPolicy: IfNotPresent
affinity:
podAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- {key: app, operator: In, values: ["redis"]}
topologyKey: kubernetes.io/hostname
EOF
解析：
1、使用硬亲和性调度，根据运行的pod标签，选择调度到相同pod的节点。
2、同时开启4个容器，测试是否调度到app=redis pod相同的节点上。
3.1.5、查询调度是否符合预期
master1 ~]# kubectl get pods -o wide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-affinity-required-5574987b9f-2rqlz 1/1 Running 0 12s 10.244.3.54 node1 <none> <none> app=pod-test,pod-template-hash=5574987b9f
pod-affinity-required-5574987b9f-c8n2j 1/1 Running 0 12s 10.244.3.56 node1 <none> <none> app=pod-test,pod-template-hash=5574987b9f
pod-affinity-required-5574987b9f-mnz44 1/1 Running 0 12s 10.244.3.57 node1 <none> <none> app=pod-test,pod-template-hash=5574987b9f
pod-affinity-required-5574987b9f-sp2qm 1/1 Running 0 12s 10.244.3.55 node1 <none> <none> app=pod-test,pod-template-hash=5574987b9f
redis-56f97b7f57-x82dj 1/1 Running 0 5m54s 10.244.3.53 node1 <none> <none> app=redis,pod-template-hash=56f97b7f57
# 可以看出来，所有都被调度到跟redis相同的节点上，硬亲和性生效
3.2、redis软亲和性-实践
3.2.1、需求
pod必须把数据存储到redis中，原则上最好在一起，实在不行的话，划分到其他节点也可以
示例：申请资源的时候cpu或内存，已经写上预期，分析软亲和性调度，是否当资源分配完成后，会不会调度到权限低的节点上。
3.2.2、定义资源配置清单且应用【创建redis的pod】
kubectl apply -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
name: redis-preferred
spec:
replicas: 1
selector:
matchLabels:
app: redis
template:
metadata:
labels:
app: redis
spec:
nodeName: node1
containers:
- name: redis
image: 192.168.10.33:80/k8s/redis:latest
imagePullPolicy: IfNotPresent
resources:
requests:
cpu: 300m
memory: 512Mi
EOF
解析：
1、提前写上预期的cpu和内存
2、使用节点调度，把它调度到node1上
3.2.3、查询redis创建的结果
master1 ~]# kubectl get pods -o wide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
redis-preferred-58975865b8-qxx8l 1/1 Running 0 32s 10.244.3.58 node1 <none> <none> app=redis,pod-template-hash=58975865b8
3.2.4、定义资源配置清单【创建测试pod，利用资源限制测试软亲和性，pod标签选择的权重调度】
kubectl apply -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
name: pod-affinity-preferred
spec:
replicas: 4
selector:
matchLabels:
app: pod-test
template:
metadata:
labels:
app: pod-test
spec:
containers:
- name: pod-test
image: 192.168.10.33:80/k8s/pod_test:v0.1
imagePullPolicy: IfNotPresent
resources:
requests:
cpu: 300m
memory: 500Mi
affinity:
podAffinity:
preferredDuringSchedulingIgnoredDuringExecution:
- weight: 100
podAffinityTerm:
labelSelector:
matchExpressions:
- {key: app, operator: In, values: ["redis"]}
topologyKey: kubernetes.io/hostname
- weight: 50
podAffinityTerm:
labelSelector:
matchExpressions:
- {key: app, operator: In, values: ["redis"]}
topologyKey: env
EOF
解析：
1、生成4个容器，每个容器cpu:300m,内存：500Mi。
2、软亲和性，pod标签匹配，按权重调度。
3、拓朴：使用标签env
3.2.5、查询运行结果且分析总结
master1 ~]# kubectl get pods -o wide --show-labels
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
pod-affinity-preferred-8447b54dd7-fcs4h 1/1 Running 0 84s 10.244.4.89 node2 <none> <none> app=pod-test,pod-template-hash=8447b54dd7
pod-affinity-preferred-8447b54dd7-gsl4x 0/1 Pending 0 84s <none> <none> <none> <none> app=pod-test,pod-template-hash=8447b54dd7
pod-affinity-preferred-8447b54dd7-jwk46 1/1 Running 0 84s 10.244.3.60 node1 <none> <none> app=pod-test,pod-template-hash=8447b54dd7
pod-affinity-preferred-8447b54dd7-xpzlf 1/1 Running 0 84s 10.244.4.90 node2 <none> <none> app=pod-test,pod-template-hash=8447b54dd7
redis-preferred-58975865b8-qxx8l 1/1 Running 0 9m35s 10.244.3.58 node1 <none> <none> app=redis,pod-template-hash=58975865b8
解析：
1、redis调度相同的节点只有一个pod
master1 ~]# kubectl describe nodes node1
...
Allocated resources:
(Total limits may be over 100 percent, i.e., overcommitted.)
Resource Requests Limits
-------- -------- ------
cpu 850m (85%) 0 (0%) # 已经使用850不够分配，所以调度到其它权重小的节点上
memory 1012Mi (27%) 0 (0%)
ephemeral-storage 0 (0%) 0 (0%)
hugepages-1Gi 0 (0%) 0 (0%)
hugepages-2Mi 0 (0%) 0 (0%)
...
2、node2为什么只调度有2个pod？
master1 ~]# kubectl describe nodes node2
Allocated resources:
(Total limits may be over 100 percent, i.e., overcommitted.)
Resource Requests Limits
-------- -------- ------
cpu 850m (85%) 0 (0%) # 因为每个容器需要占用300，已经到850，无法在分配，所以一直挂起，直到有节点符合资源配置，才被调度运行
memory 1000Mi (27%) 0 (0%)
ephemeral-storage 0 (0%) 0 (0%)
hugepages-1Gi 0 (0%) 0 (0%)
hugepages-2Mi 0 (0%) 0 (0%)
3、查看panding原因
master1 ~]# kubectl describe pod pod-affinity-preferred-8447b54dd7-gsl4x
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning FailedScheduling 68s default-scheduler 0/5 nodes are available: 2 Insufficient cpu, 3 node(s)
had untolerated taint {node-role.kubernetes.io/control-plane: }. preemption: 0/5 nodes are available: 2 No
preemption victims found for incoming pod, 3 Preemption is not helpful for scheduling.
# 从上面日志分析，就是CPU资源不够分配的原因。
浙公网安备 33010602011771号