前言
- 本篇来学习Pod的调度
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: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。
创建一个pod-nodename.yaml文件,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-nodename namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 nodeName: node1 # 指定调度到node1节点上
# 创建pod [root@master ~]# kubectl create -f pod-nodename.yaml pod/pod-nodename created #查看Pod调度到NODE属性,确实是调度到了node1节点上 [root@master ~]# kubectl get pods pod-nodename -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodename 1/1 Running 0 6s 10.244.1.28 node1 <none> <none> # 删除原来pod, [root@master ~]# kubectl delete -f pod-nodename.yaml pod "pod-nodename" deleted # 修改nodeName值为node4(不存在的节点名称) [root@master ~]# vim pod-nodename.yaml # 重新创建pod [root@master ~]# kubectl create -f pod-nodename.yaml pod/pod-nodename created #再次查看,发现已经向Node4节点调度,但是由于不存在node4节点,所以pod无法正常运行,处于Pending状态 NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodename 0/1 Pending 0 7s <none> node4 <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
[root@master ~]# kubectl label nodes node1 nodeenv=pro node/node1 labeled [root@master ~]# kubectl label nodes node2 nodeenv=test node/node2 labeled
2 创建一个pod-nodeselector.yaml文件,并使用它创建Pod
apiVersion: v1 kind: Pod metadata: name: pod-nodeselector namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 nodeSelector: nodeenv: pro # 指定调度到具有nodeenv=pro标签的节点上
#创建Pod [root@master ~]# kubectl create -f pod-nodeselector.yaml pod/pod-nodeselector created # 查看Pod调度到NODE属性,确实是调度到了node1节点上 [root@master ~]# kubectl get pods pod-nodeselector -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodeselector 1/1 Running 0 76s 10.244.1.29 node1 <none> <none> # 删除pod [root@master ~]# kubectl delete -f pod-nodeselector.yaml pod "pod-nodeselector" deleted # 修改 pod-nodeselector.yaml 中nodeenv值为pre(不存在打此标签的节点) [root@master ~]# vim pod-nodeselector.yaml # 重新创建pod [root@master ~]# kubectl create -f pod-nodeselector.yaml pod/pod-nodeselector created #再次查看,发现pod无法正常运行,Node的值为none [root@master ~]# kubectl get pod pod-nodeselector -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodeselector 0/1 Pending 0 33s <none> <none> <none> <none> # 查看详情,发现node selector匹配失败的提示 [root@master ~]# kubectl describe pods pod-nodeselector -n dev Name: pod-nodeselector Namespace: dev Events: Type Reason Age From Message ---- ------ ---- ---- ------- Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 4 node(s) didn't match node selector. Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 4 node(s) didn't match node selector.
亲和性调度
上面,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:
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 operat or 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。
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,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-nodeaffinity-required namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 affinity: #亲和性设置 nodeAffinity: #设置node亲和性 requiredDuringSchedulingIgnoredDuringExecution: # 硬限制 nodeSelectorTerms: - matchExpressions: # 匹配env的值在["xxx","yyy"]中的标签 - key: nodeenv operator: In values: ["xxx","yyy"]
# 创建pod [root@master ~]# kubectl create -f pod-nodeaffinity-required.yaml pod/pod-nodeaffinity-required created # 查看pod状态 (运行失败) [root@master ~]# kubectl get pods pod-nodeaffinity-required -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodeaffinity-required 0/1 Pending 0 25m <none> <none> <none> <none> # 查看Pod的详情 发现调度失败,提示node选择失败 [root@master ~]# kubectl describe pod pod-nodeaffinity-required -n dev Name: pod-nodeaffinity-required Namespace: dev Events: Type Reason Age From Message ---- ------ ---- ---- ------- Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 4 node(s) didn't match node selector. Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 4 node(s) didn't match node selector. #接下来,停止pod [root@master ~]# kubectl delete -f pod-nodeaffinity-required.yaml pod "pod-nodeaffinity-required" deleted # 修改文件,将values: ["xxx","yyy"]------> ["test","yyy"] [root@master ~]# vim pod-nodeaffinity-required.yaml # 再次启动 [root@master ~]# kubectl create -f pod-nodeaffinity-required.yaml pod/pod-nodeaffinity-required created # 此时查看,发现调度成功,已经将pod调度到了node2上 [root@master ~]# kubectl get pods pod-nodeaffinity-required -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-nodeaffinity-required 1/1 Running 0 16s 10.244.3.20 node2 <none> <none>
接下来再演示一下requiredDuringSchedulingIgnoredDuringExecution
创建pod-nodeaffinity-preferred.yaml,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-nodeaffinity-preferred namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 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"]
# 创建pod [root@master ~]# kubectl create -f pod-nodeaffinity-preferred.yaml pod/pod-nodeaffinity-preferred created # 查看pod状态 (运行成功) [root@master ~]# kubectl get pod pod-nodeaffinity-preferred -n dev NAME READY STATUS RESTARTS AGE pod-nodeaffinity-preferred 1/1 Running 0 6s
NodeAffinity规则设置的注意事项:
- 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
- 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
- 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
- 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:
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
,
- 首先创建一个参照Pod,pod-podaffinity-target.yaml,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-podaffinity-target namespace: dev labels: podenv: pro #设置标签 spec: containers: - name: nginx image: nginx:1.17.1 nodeName: node1 # 将目标pod名确指定到node1上
# 创建pod [root@master ~]# kubectl create -f pod-podaffinity-target.yaml pod/pod-podaffinity-target created # 查看pod运行状态 [root@master ~]# kubectl get pods pod-podaffinity-target -n dev NAME READY STATUS RESTARTS AGE pod-podaffinity-target 1/1 Running 0 6s
- 创建pod-podaffinity-required.yaml,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-podaffinity-required namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 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。
# 启动pod [root@master ~]# vim pod-podaffinity-required.yaml [root@master ~]# kubectl create -f pod-podaffinity-required.yaml pod/pod-podaffinity-required created # 查看pod状态,发现未运行 [root@master ~]# kubectl get pods pod-podaffinity-required -n dev NAME READY STATUS RESTARTS AGE pod-podaffinity-required 0/1 Pending 0 6s # 查看详细信息 [root@master ~]# kubectl describe pods pod-podaffinity-required -n dev Name: pod-podaffinity-required Namespace: dev Events: Type Reason Age From Message ---- ------ ---- ---- ------- Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 1 node(s) had taints that the pod didn't tolerate, 3 node(s) didn't match pod affinity rules. Warning FailedScheduling <unknown> default-scheduler 0/4 nodes are available: 1 node(s) had taints that the pod didn't tolerate, 3 node(s) didn't match pod affinity rules. # 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 [root@master ~]# vim pod-podaffinity-required.yaml # 然后重新创建pod,查看效果 [root@master ~]# kubectl delete -f pod-podaffinity-required.yaml pod "pod-podaffinity-required" deleted [root@master ~]# kubectl create -f pod-podaffinity-required.yaml pod/pod-podaffinity-required created # 发现此时Pod运行正常 [root@master ~]# kubectl get pods pod-podaffinity-required -n dev NAME READY STATUS RESTARTS AGE pod-podaffinity-required 1/1 Running 0 8s
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。
- 继续使用上个案例中目标pod
# 查看pod [root@master ~]# kubectl get pods pod-podaffinity-target -n dev -o wide --show-labels NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELS pod-podaffinity-target 1/1 Running 0 13m 10.244.1.30 node1 <none> <none> podenv=pro
- 创建pod-podantiaffinity-required.yaml,内容如下:
apiVersion: v1 kind: Pod metadata: name: pod-podantiaffinity-required namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 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。
# 创建pod [root@master ~]# kubectl create -f pod-podantiaffinity-required.yaml pod/pod-podantiaffinity-required created # 查看pod 发现调度到了node2上 [root@master ~]# kubectl get pods pod-podantiaffinity-required -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-podantiaffinity-required 1/1 Running 0 10s 10.244.3.22 node2 <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:
# 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
- 准备节点node1(为了演示效果更加明显,只保留node1节点,暂时停止其它点)
- 为node1节点设置一个污点:
tag=heima:PreferNoSchedule
;然后创建pod1( pod1 可以 ) - 修改为node1节点设置一个污点:
tag=heima:NoSchedule
;然后创建pod2( pod1 正常 pod2 失败 ) - 修改为node1节点设置一个污点:
tag=heima:NoExecute
;然后创建pod3 ( 3个pod都失败 )
# 为node1设置污点(PreferNoSchedule) [root@master ~]# kubectl taint nodes node1 tag=heima:PreferNoSchedule node/node1 tainted # 创建pod1 [root@master ~]# kubectl run taint1 --image=nginx:1.17.1 -n dev kubectl run --generator=deployment/apps.v1 is DEPRECATED and will be removed in a future version. Use kubectl run --generator=run-pod/v1 or kubectl create instead. deployment.apps/taint1 created # 查看pod [root@master ~]# kubectl get pods -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES taint1-766c47bf55-gq9xc 1/1 Running 0 7s 10.244.1.34 node1 <none> <none> # 为node1设置污点(取消PreferNoSchedule,设置NoSchedule) [root@master ~]# kubectl taint nodes node1 tag:PreferNoSchedule- node/node1 untainted [root@master ~]# kubectl taint nodes node1 tag=heima:NoSchedule node/node1 tainted # 创建pod2 [root@master ~]# kubectl run taint2 --image=nginx:1.17.1 -n dev kubectl run --generator=deployment/apps.v1 is DEPRECATED and will be removed in a future version. Use kubectl run --generator=run-pod/v1 or kubectl create instead. deployment.apps/taint2 created # 查看pod [root@master ~]# kubectl get pods -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES taint1-766c47bf55-gq9xc 1/1 Running 0 5m12s 10.244.1.34 node1 <none> <none> taint2-84946958cf-tb54s 0/1 Pending 0 73s <none> <none> <none> <none> # 为node1设置污点(取消NoSchedule,设置NoExecute) [root@master ~]# kubectl taint nodes node1 tag:NoSchedule- node/node1 untainted [root@master ~]# kubectl taint nodes node1 tag=heima:NoExecute node/node1 tainted # 创建pod3 [root@master ~]# kubectl run taint3 --image=nginx:1.17.1 -n dev kubectl run --generator=deployment/apps.v1 is DEPRECATED and will be removed in a future version. Use kubectl run --generator=run-pod/v1 or kubectl create instead. deployment.apps/taint3 created # 查看pod [root@master ~]# kubectl get pods -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES taint1-766c47bf55-dbc8d 0/1 Pending 0 17s <none> <none> <none> <none> taint2-84946958cf-cgzwp 0/1 Pending 0 17s <none> <none> <none> <none> taint3-57d45f9d4c-69tzx 0/1 Pending 0 9s <none> <none> <none> <none>
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
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。
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:
- 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 - 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:
apiVersion: v1 kind: Pod metadata: name: pod-toleration namespace: dev spec: containers: - name: nginx image: nginx:1.17.1
# 创建pod [root@master ~]# kubectl create -f pod-toleration.yaml pod/pod-toleration created # 未添加容忍前,查看pod [root@master ~]# kubectl get pods -n dev -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-toleration 0/1 Pending 0 8s <none> <none> <none> <none>
修改pod-toleration.yamlyaml, 添加容忍
apiVersion: v1 kind: Pod metadata: name: pod-toleration namespace: dev spec: containers: - name: nginx image: nginx:1.17.1 tolerations: # 添加容忍 - key: "tag" # 要容忍的污点的key operator: "Equal" # 操作符 value: "heima" # 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
# 删除pod [root@master ~]# kubectl delete -f pod-toleration.yaml pod "pod-toleration" deleted # 重新启动pod [root@master ~]# kubectl create -f pod-toleration.yaml pod/pod-toleration created # 添加容忍后查看pod [root@master ~]# kubectl get pod -n dev NAME READY STATUS RESTARTS AGE pod-toleration 1/1 Running 0 8s