08-27-周三_17-09-29
This commit is contained in:
552
Kubernetes/调度器.md
Normal file
552
Kubernetes/调度器.md
Normal file
@@ -0,0 +1,552 @@
|
||||
# Pod调度
|
||||
|
||||
在默认情况下，一个Pod在哪个Node节点上运行，是由Scheduler组件采用相应的算法计算出来的，这个过程是不受人工控制的。但是在实际使用中，这并不满足的需求，因为很多情况下，我们想控制某些Pod到达某些节点上，那么应该怎么做呢？这就要求了解kubernetes对Pod的调度规则，kubernetes提供了四大类调度方式：
|
||||
|
||||
- 自动调度：运行在哪个节点上完全由Scheduler经过一系列的算法计算得出
|
||||
- 定向调度：NodeName、NodeSelector
|
||||
- 亲和性调度：NodeAffinity、PodAffinity、PodAntiAffinity
|
||||
- 污点（容忍）调度：Taints、Toleration
|
||||
|
||||
## 定向调度
|
||||
|
||||
定向调度，指的是利用在pod上声明nodeName或者nodeSelector，以此将Pod调度到期望的node节点上。注意，这里的调度是强制的，这就意味着即使要调度的目标Node不存在，也会向上面进行调度，只不过pod运行失败而已。
|
||||
|
||||
### **NodeName**
|
||||
|
||||
NodeName用于强制约束将Pod调度到指定的Name的Node节点上。这种方式，其实是直接跳过Scheduler的调度逻辑，直接将Pod调度到指定名称的节点。
|
||||
|
||||
接下来，实验一下：创建一个pod-nodename.yaml文件
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-nodename
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
nodeName: node01 # 指定调度到nodem01节点上
|
||||
```
|
||||
|
||||
```shell
|
||||
#查看Pod调度到NODE属性，确实是调度到了node01节点上
|
||||
$ kubectl get pods pod-nodename -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodename 1/1 Running 0 62s 10.244.196.188 node01 <none> <none>
|
||||
|
||||
# 接下来，删除pod，修改nodeName的值为node03（并没有node03节点）
|
||||
#再次查看，发现已经向Node03节点调度，但是由于不存在node03节点，所以pod无法正常运行
|
||||
$ kubectl get pods pod-nodename -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodename 0/1 Pending 0 10s <none> node03 <none> <none>
|
||||
```
|
||||
|
||||
### **NodeSelector**
|
||||
|
||||
NodeSelector用于将pod调度到添加了指定标签的node节点上。它是通过kubernetes的label-selector机制实现的，也就是说，在pod创建之前，会由scheduler使用MatchNodeSelector调度策略进行label匹配，找出目标node，然后将pod调度到目标节点，该匹配规则是强制约束。
|
||||
|
||||
接下来，实验一下：
|
||||
|
||||
1 首先分别为node节点添加标签
|
||||
|
||||
```shell
|
||||
$ kubectl label nodes node01 nodeenv=pro
|
||||
node/node02 labeled
|
||||
$ kubectl label nodes node02 nodeenv=test
|
||||
node/node02 labeled
|
||||
```
|
||||
|
||||
2 创建一个pod-nodeselector.yaml文件，并使用它创建Pod
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-nodeselector
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
nodeSelector:
|
||||
nodeenv: pro # 指定调度到具有nodeenv=pro标签的节点上
|
||||
```
|
||||
|
||||
```shell
|
||||
#查看Pod调度到NODE属性，确实是调度到了node1节点上
|
||||
$ kubectl get pod pod-nodeselector -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodeselector 1/1 Running 0 14s 10.244.196.189 node01 <none> <none>
|
||||
|
||||
# 接下来，删除pod，修改nodeSelector的值为nodeenv: xxxx（不存在打有此标签的节点）
|
||||
#再次查看，发现pod无法正常运行,Node的值为none
|
||||
$ kubectl get pod -o wide
|
||||
Events:
|
||||
Type Reason Age From Message
|
||||
---- ------ ---- ---- -------
|
||||
Warning FailedScheduling 36s default-scheduler 0/3 nodes are available: 1 node(s) had untolerated taint {node-role.kubernetes.io/control-plane: }, 2 node(s) didn't match Pod's node affinity/selector. preemption: 0/3 nodes are available: 3 Preemption is not helpful for scheduling.
|
||||
```
|
||||
|
||||
## 亲和性调度
|
||||
|
||||
上一节，介绍了两种定向调度的方式，使用起来非常方便，但是也有一定的问题，那就是如果没有满足条件的Node，那么Pod将不会被运行，即使在集群中还有可用Node列表也不行，这就限制了它的使用场景。
|
||||
|
||||
基于上面的问题，kubernetes还提供了一种亲和性调度（Affinity）。它在**NodeSelector**的基础之上的进行了扩展，可以通过配置的形式，实现优先选择满足条件的Node进行调度，如果没有，也可以调度到不满足条件的节点上，使调度更加灵活。
|
||||
|
||||
Affinity主要分为三类：
|
||||
|
||||
- nodeAffinity(node亲和性）: 以node为目标，解决pod可以调度到哪些node的问题
|
||||
- podAffinity(pod亲和性) : 以pod为目标，解决pod可以和哪些已存在的pod部署在同一个拓扑域中的问题
|
||||
- podAntiAffinity(pod反亲和性) : 以pod为目标，解决pod不能和哪些已存在pod部署在同一个拓扑域中的问题
|
||||
|
||||
> 关于亲和性(反亲和性)使用场景的说明：
|
||||
>
|
||||
> **亲和性**：如果两个应用频繁交互，那就有必要利用亲和性让两个应用的尽可能的靠近，这样可以减少因网络通信而带来的性能损耗。
|
||||
>
|
||||
> **反亲和性**：当应用的采用多副本部署时，有必要采用反亲和性让各个应用实例打散分布在各个node上，这样可以提高服务的高可用性。
|
||||
|
||||
### **NodeAffinity**
|
||||
|
||||
首先来看一下`NodeAffinity`的可配置项：
|
||||
|
||||
```shell
|
||||
pod.spec.affinity.nodeAffinity
|
||||
requiredDuringSchedulingIgnoredDuringExecution Node节点必须满足指定的所有规则才可以，相当于硬限制
|
||||
nodeSelectorTerms 节点选择列表
|
||||
matchFields 按节点字段列出的节点选择器要求列表
|
||||
matchExpressions 按节点标签列出的节点选择器要求列表(推荐)
|
||||
key 键
|
||||
values 值
|
||||
operator 关系符 支持Exists, DoesNotExist, In, NotIn, Gt, Lt
|
||||
preferredDuringSchedulingIgnoredDuringExecution 优先调度到满足指定的规则的Node，相当于软限制 (倾向)
|
||||
preference 一个节点选择器项，与相应的权重相关联
|
||||
matchFields 按节点字段列出的节点选择器要求列表
|
||||
matchExpressions 按节点标签列出的节点选择器要求列表(推荐)
|
||||
key 键
|
||||
values 值
|
||||
operator 关系符 支持In, NotIn, Exists, DoesNotExist, Gt, Lt
|
||||
weight 倾向权重，在范围1-100。
|
||||
```
|
||||
|
||||
```shell
|
||||
关系符的使用说明:
|
||||
|
||||
- matchExpressions:
|
||||
- key: nodeenv # 匹配存在标签的key为nodeenv的节点
|
||||
operator: Exists
|
||||
- key: nodeenv # 匹配标签的key为nodeenv,且value是"xxx"或"yyy"的节点
|
||||
operator: In
|
||||
values: ["xxx","yyy"]
|
||||
- key: nodeenv # 匹配标签的key为nodeenv,且value大于"xxx"的节点
|
||||
operator: Gt
|
||||
values: "xxx"
|
||||
```
|
||||
|
||||
接下来首先演示一下`requiredDuringSchedulingIgnoredDuringExecution` ,
|
||||
|
||||
创建pod-nodeaffinity-required.yaml
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-nodeaffinity-required
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
affinity: #亲和性设置
|
||||
nodeAffinity: #设置node亲和性
|
||||
requiredDuringSchedulingIgnoredDuringExecution: # 硬限制
|
||||
nodeSelectorTerms:
|
||||
- matchExpressions: # 匹配nodeenv的值在["xxx","yyy"]中的标签
|
||||
- key: nodeenv
|
||||
operator: In
|
||||
values: ["xxx","yyy"]
|
||||
```
|
||||
|
||||
```shell
|
||||
# 查看pod状态 （运行失败）
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodeaffinity-required 0/1 Pending 0 6s <none> <none> <none> <none>
|
||||
|
||||
# 查看Pod的详情
|
||||
# 发现调度失败，提示node选择失败
|
||||
$ kubectl describe pod pod-nodeaffinity-required
|
||||
Events:
|
||||
Type Reason Age From Message
|
||||
---- ------ ---- ---- -------
|
||||
Warning FailedScheduling 42s default-scheduler 0/3 nodes are available: 1 node(s) had untolerated taint {node-role.kubernetes.io/control-plane: }, 2 node(s) didn't match Pod's node affinity/selector. preemption: 0/3 nodes are available: 3 Preemption is not helpful for scheduling.
|
||||
|
||||
# 修改文件，将values: ["xxx","yyy"]------> ["pro","yyy"]
|
||||
# 此时查看，发现调度成功，已经将pod调度到了node1上
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodeaffinity-required 1/1 Running 0 7s 10.244.196.187 node01 <none> <none>
|
||||
```
|
||||
|
||||
接下来再演示一下`preferredDuringSchedulingIgnoredDuringExecution` ,
|
||||
|
||||
创建pod-nodeaffinity-preferred.yaml
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-nodeaffinity-preferred
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
affinity: #亲和性设置
|
||||
nodeAffinity: #设置node亲和性
|
||||
preferredDuringSchedulingIgnoredDuringExecution: # 软限制
|
||||
- weight: 1
|
||||
preference:
|
||||
matchExpressions: # 匹配env的值在["xxx","yyy"]中的标签(当前环境没有)
|
||||
- key: nodeenv
|
||||
operator: In
|
||||
values: ["xxx","yyy"]
|
||||
```
|
||||
|
||||
```shell
|
||||
# 创建pod
|
||||
[root@k8s-master01 ~]# kubectl create -f pod-nodeaffinity-preferred.yaml
|
||||
pod/pod-nodeaffinity-preferred created
|
||||
|
||||
# 查看pod状态 （运行成功）
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-nodeaffinity-preferred 1/1 Running 0 6s 10.244.196.186 node01 <none> <none>
|
||||
```
|
||||
|
||||
```
|
||||
NodeAffinity规则设置的注意事项：
|
||||
1 如果同时定义了nodeSelector和nodeAffinity，那么必须两个条件都得到满足，Pod才能运行在指定的Node上
|
||||
2 如果nodeAffinity指定了多个nodeSelectorTerms，那么只需要其中一个能够匹配成功即可
|
||||
3 如果一个nodeSelectorTerms中有多个matchExpressions ，则一个节点必须满足所有的才能匹配成功
|
||||
4 如果一个pod所在的Node在Pod运行期间其标签发生了改变，不再符合该Pod的节点亲和性需求，则系统将忽略此变化
|
||||
```
|
||||
|
||||
### **PodAffinity**
|
||||
|
||||
PodAffinity主要实现以运行的Pod为参照，实现让新创建的Pod跟参照pod在一个区域的功能。
|
||||
|
||||
首先来看一下`PodAffinity`的可配置项：
|
||||
|
||||
```shell
|
||||
pod.spec.affinity.podAffinity
|
||||
requiredDuringSchedulingIgnoredDuringExecution 硬限制
|
||||
namespaces 指定参照pod的namespace
|
||||
topologyKey 指定调度作用域
|
||||
labelSelector 标签选择器
|
||||
matchExpressions 按节点标签列出的节点选择器要求列表(推荐)
|
||||
key 键
|
||||
values 值
|
||||
operator 关系符 支持In, NotIn, Exists, DoesNotExist.
|
||||
matchLabels 指多个matchExpressions映射的内容
|
||||
preferredDuringSchedulingIgnoredDuringExecution 软限制
|
||||
podAffinityTerm 选项
|
||||
namespaces
|
||||
topologyKey
|
||||
labelSelector
|
||||
matchExpressions
|
||||
key 键
|
||||
values 值
|
||||
operator
|
||||
matchLabels
|
||||
weight 倾向权重，在范围1-100
|
||||
```
|
||||
|
||||
```
|
||||
topologyKey用于指定调度时作用域,例如:
|
||||
如果指定为kubernetes.io/hostname，那就是以Node节点为区分范围
|
||||
如果指定为beta.kubernetes.io/os,则以Node节点的操作系统类型来区分
|
||||
```
|
||||
|
||||
接下来，演示下`requiredDuringSchedulingIgnoredDuringExecution`,
|
||||
|
||||
1）首先创建一个参照Pod，pod-podaffinity-target.yaml：
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-podaffinity-target
|
||||
labels:
|
||||
podenv: pro #设置标签
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
nodeName: node01 # 将目标pod名确指定到node01上
|
||||
```
|
||||
|
||||
```shell
|
||||
# 查看pod状况
|
||||
$ kubectl get pod -o wide --show-labels
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS
|
||||
pod-podaffinity-target 1/1 Running 0 10s 10.244.196.191 node01 <none> <none> podenv=pro
|
||||
```
|
||||
|
||||
2）创建pod-podaffinity-required.yaml，内容如下：
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-podaffinity-required
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
affinity: #亲和性设置
|
||||
podAffinity: #设置pod亲和性
|
||||
requiredDuringSchedulingIgnoredDuringExecution: # 硬限制
|
||||
- labelSelector:
|
||||
matchExpressions: # 匹配env的值在["xxx","yyy"]中的标签
|
||||
- key: podenv
|
||||
operator: In
|
||||
values: ["xxx","yyy"]
|
||||
topologyKey: kubernetes.io/hostname
|
||||
```
|
||||
|
||||
上面配置表达的意思是：新Pod必须要与拥有标签nodeenv=xxx或者nodeenv=yyy的pod在同一Node上，显然现在没有这样pod，接下来，运行测试一下。
|
||||
|
||||
```shell
|
||||
# 查看pod状态，发现未运行
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-podaffinity-required 0/1 Pending 0 67s <none> <none> <none> <none>
|
||||
pod-podaffinity-target 1/1 Running 0 2m29s 10.244.196.191 node01 <none> <none>
|
||||
|
||||
# 查看详细信息
|
||||
$ kubectl describe pod pod-podaffinity-required
|
||||
Events:
|
||||
Type Reason Age From Message
|
||||
---- ------ ---- ---- -------
|
||||
Warning FailedScheduling 106s default-scheduler 0/3 nodes are available: 1 node(s) had untolerated taint {node-role.kubernetes.io/control-plane: }, 2 node(s) didn't match pod affinity rules. preemption: 0/3 nodes are available: 3 Preemption is not helpful for scheduling.
|
||||
|
||||
# 接下来修改 values: ["xxx","yyy"]----->values:["pro","yyy"]
|
||||
# 意思是：新Pod必须要与拥有标签nodeenv=xxx或者nodeenv=yyy的pod在同一Node上
|
||||
# 然后重新创建pod，查看效果
|
||||
# 发现此时Pod运行正常
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-podaffinity-required 1/1 Running 0 15s 10.244.196.129 node01 <none> <none>
|
||||
pod-podaffinity-target 1/1 Running 0 4m8s 10.244.196.191 node01 <none> <none>
|
||||
```
|
||||
|
||||
关于`PodAffinity`的 `preferredDuringSchedulingIgnoredDuringExecution`，这里不再演示。
|
||||
|
||||
### **PodAntiAffinity**
|
||||
|
||||
PodAntiAffinity主要实现以运行的Pod为参照，让新创建的Pod跟参照pod不在一个区域中的功能。
|
||||
|
||||
它的配置方式和选项跟PodAffinty是一样的，这里不再做详细解释，直接做一个测试案例。
|
||||
|
||||
1）继续使用上个案例中目标pod
|
||||
|
||||
```shell
|
||||
$ kubectl get pod --show-labels
|
||||
NAME READY STATUS RESTARTS AGE LABELS
|
||||
pod-podaffinity-target 1/1 Running 0 5m31s podenv=pro
|
||||
```
|
||||
|
||||
2）创建pod-podantiaffinity-required.yaml，内容如下：
|
||||
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Pod
|
||||
metadata:
|
||||
name: pod-podantiaffinity-required
|
||||
spec:
|
||||
containers:
|
||||
- name: nginx
|
||||
image: aaronxudocker/myapp:v1.0
|
||||
affinity: #亲和性设置
|
||||
podAntiAffinity: #设置pod亲和性
|
||||
requiredDuringSchedulingIgnoredDuringExecution: # 硬限制
|
||||
- labelSelector:
|
||||
matchExpressions: # 匹配podenv的值在["pro"]中的标签
|
||||
- key: podenv
|
||||
operator: In
|
||||
values: ["pro"]
|
||||
topologyKey: kubernetes.io/hostname
|
||||
```
|
||||
|
||||
上面配置表达的意思是：新Pod必须要与拥有标签nodeenv=pro的pod不在同一Node上，运行测试一下。
|
||||
|
||||
```shell
|
||||
# 查看pod
|
||||
# 发现调度到了node2上
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
pod-podaffinity-target 1/1 Running 0 6m30s 10.244.196.191 node01 <none> <none>
|
||||
pod-podantiaffinity-required 1/1 Running 0 8s 10.244.140.122 node02 <none> <none>
|
||||
```
|
||||
|
||||
## 污点和容忍
|
||||
|
||||
**污点（Taints）**
|
||||
|
||||
前面的调度方式都是站在Pod的角度上，通过在Pod上添加属性，来确定Pod是否要调度到指定的Node上，其实我们也可以站在Node的角度上，通过在Node上添加**污点**属性，来决定是否允许Pod调度过来。
|
||||
|
||||
Node被设置上污点之后就和Pod之间存在了一种相斥的关系，进而拒绝Pod调度进来，甚至可以将已经存在的Pod驱逐出去。
|
||||
|
||||
污点的格式为：`key=value:effect`, key和value是污点的标签，effect描述污点的作用，支持如下三个选项：
|
||||
|
||||
- PreferNoSchedule：kubernetes将**尽量避免把Pod调度到具有该污点的Node上**，除非没有其他节点可调度
|
||||
- NoSchedule：kubernetes将**不会把Pod调度到具有该污点的Node上**，**但不会影响当前Node上已存在的Pod**
|
||||
- NoExecute：kubernetes将**不会把Pod调度到具有该污点的Node上**，**同时也会将Node上已存在的Pod驱离**
|
||||
|
||||
|
||||
|
||||
使用kubectl设置和去除污点的命令示例如下：
|
||||
|
||||
```shell
|
||||
# 设置污点
|
||||
kubectl taint nodes node1 key=value:effect
|
||||
|
||||
# 去除污点
|
||||
kubectl taint nodes node1 key:effect-
|
||||
|
||||
# 去除所有污点
|
||||
kubectl taint nodes node1 key-
|
||||
```
|
||||
|
||||
接下来，演示下污点的效果：
|
||||
|
||||
```shell
|
||||
# 为node01设置污点(PreferNoSchedule)
|
||||
$ kubectl taint node node01 tag=eagle:PreferNoSchedule
|
||||
node/node01 tainted
|
||||
|
||||
# 创建100个pod，全部都运行在了node02节点上
|
||||
$ kubectl create deployment myapp --image=aaronxudocker/myapp:v1.0 --replicas=100
|
||||
deployment.apps/myapp created
|
||||
$ kubectl get pod -o wide |grep node02|wc -l
|
||||
100
|
||||
|
||||
# 创建1000个pod的时候，node02节点性能不足，不满足预选条件，部分pod运行到了node01上面
|
||||
$ kubectl get pod -o wide |grep node01|wc -l
|
||||
106
|
||||
$ kubectl get pod -o wide |grep node02|wc -l
|
||||
103
|
||||
|
||||
# 为node1设置污点(取消PreferNoSchedule，设置NoSchedule)
|
||||
$ kubectl taint nodes node01 tag:PreferNoSchedule-
|
||||
node/node01 untainted
|
||||
$ kubectl taint nodes node01 tag=eagle:NoSchedule
|
||||
node/node01 tainted
|
||||
|
||||
# 创建pod
|
||||
$ kubectl create deployment myapp --image=aaronxudocker/myapp:v1.0 --replicas=100
|
||||
deployment.apps/myapp created
|
||||
|
||||
$ kubectl get pod -o wide | grep node02 | wc -l
|
||||
100
|
||||
|
||||
# 关闭node02虚拟机后，再次创建，无法被调度
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
myapp-9d57f8b4-2jf2f 0/1 Pending 0 4s <none> <none> <none> <none>
|
||||
myapp-9d57f8b4-2rkhr 0/1 Pending 0 8s <none> <none> <none> <none>
|
||||
myapp-9d57f8b4-2sfbn 0/1 Pending 0 7s <none> <none> <none> <none>
|
||||
myapp-9d57f8b4-44hjs 0/1 Pending 0 4s <none> <none> <none> <none>
|
||||
|
||||
|
||||
# 为node1设置污点(取消NoSchedule，设置NoExecute)
|
||||
$ kubectl taint nodes node01 tag:NoSchedule-
|
||||
|
||||
# 创建10个pod，里面有部分是在node01上运行的
|
||||
$ kubectl create deployment myapp --image=aaronxudocker/myapp:v1.0 --replicas=10
|
||||
deployment.apps/myapp created
|
||||
|
||||
$ kubectl get pod -o wide |grep node01 |wc -l
|
||||
5
|
||||
|
||||
# 设置NoExecute，所有的pod运行到了node02上
|
||||
$ kubectl taint nodes node01 tag=eagle:NoExecute
|
||||
$ kubectl get pod -o wide |grep node01 |wc -l
|
||||
0
|
||||
```
|
||||
|
||||
```
|
||||
小提示：
|
||||
使用kubeadm搭建的集群，默认就会给master节点添加一个污点标记,所以pod就不会调度到master节点上.
|
||||
```
|
||||
|
||||
**容忍（Toleration）**
|
||||
|
||||
上面介绍了污点的作用，我们可以在node上添加污点用于拒绝pod调度上来，但是如果就是想将一个pod调度到一个有污点的node上去，这时候应该怎么做呢？这就要使用到**容忍**。
|
||||
|
||||
<img src="调度器/image-20240918103742892.png" alt="image-20240918103742892" style="zoom:33%;" />
|
||||
|
||||
> 污点就是拒绝，容忍就是忽略，Node通过污点拒绝pod调度上去，Pod通过容忍忽略拒绝
|
||||
|
||||
下面先通过一个案例看下效果：
|
||||
|
||||
1. 上一小节，已经在node1节点上打上了`NoExecute`的污点，此时pod是调度不上去的
|
||||
2. 本小节，可以通过给pod添加容忍，然后将其调度上去
|
||||
|
||||
创建pod-toleration.yaml,内容如下
|
||||
|
||||
```yaml
|
||||
apiVersion: apps/v1
|
||||
kind: Deployment
|
||||
metadata:
|
||||
labels:
|
||||
app: myapp
|
||||
name: myapp
|
||||
spec:
|
||||
replicas: 10
|
||||
selector:
|
||||
matchLabels:
|
||||
app: myapp
|
||||
template:
|
||||
metadata:
|
||||
labels:
|
||||
app: myapp
|
||||
spec:
|
||||
containers:
|
||||
- image: aaronxudocker/myapp:v1.0
|
||||
name: myapp
|
||||
tolerations: # 添加容忍
|
||||
- key: "tag" # 要容忍的污点的key
|
||||
operator: "Equal" # 操作符
|
||||
value: "eagle" # 容忍的污点的value
|
||||
effect: "NoExecute" # 添加容忍的规则，这里必须和标记的污点规则相同
|
||||
```
|
||||
|
||||
```shell
|
||||
# 添加容忍之后的pod
|
||||
$ kubectl get pod -o wide
|
||||
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
|
||||
myapp-6687b4f9d7-562lv 1/1 Running 0 3s 10.244.140.150 node02 <none> <none>
|
||||
myapp-6687b4f9d7-6m7wb 1/1 Running 0 3s 10.244.196.128 node01 <none> <none>
|
||||
myapp-6687b4f9d7-d9tgk 1/1 Running 0 3s 10.244.196.168 node01 <none> <none>
|
||||
myapp-6687b4f9d7-fsqw7 1/1 Running 0 3s 10.244.140.159 node02 <none> <none>
|
||||
myapp-6687b4f9d7-j7c4x 1/1 Running 0 3s 10.244.140.158 node02 <none> <none>
|
||||
myapp-6687b4f9d7-j7dfx 1/1 Running 0 3s 10.244.196.178 node01 <none> <none>
|
||||
myapp-6687b4f9d7-ll6z2 1/1 Running 0 3s 10.244.196.140 node01 <none> <none>
|
||||
myapp-6687b4f9d7-lwqj4 1/1 Running 0 3s 10.244.140.167 node02 <none> <none>
|
||||
myapp-6687b4f9d7-mvdtg 1/1 Running 0 3s 10.244.196.167 node01 <none> <none>
|
||||
myapp-6687b4f9d7-qlw5x 1/1 Running 0 3s 10.244.140.147 node02 <none> <none>
|
||||
```
|
||||
|
||||
下面看一下容忍的详细配置:
|
||||
|
||||
```shell
|
||||
$ kubectl explain pod.spec.tolerations
|
||||
......
|
||||
FIELDS:
|
||||
key # 对应着要容忍的污点的键，空意味着匹配所有的键
|
||||
value # 对应着要容忍的污点的值
|
||||
operator # key-value的运算符，支持Equal和Exists（默认）
|
||||
effect # 对应污点的effect，空意味着匹配所有影响
|
||||
tolerationSeconds # 容忍时间, 当effect为NoExecute时生效，表示pod在Node上的停留时间
|
||||
```
|
||||
|
||||
|
||||
Reference in New Issue
Block a user