Kubernetes守护进程集 DaemonSet

守护进程集 DaemonSet

守护进程集也有副本数概念,但是副本数概念并不是通过配置清单的方式人为去定义的,他是靠你当前集群的节点个数,比如我当前集群

之前安装集群安装的网络插件calico

[root@k8s-master1 deployment]# kubectl get pods -n kube-system
NAME                                             READY   STATUS    RESTARTS       AGE
calico-kube-controllers-6bd6b69df9-rl2mh         1/1     Running   21 (57m ago)   11d
calico-node-52wx8                                1/1     Running   29 (61m ago)   11d
#这个calico 网络插件的 个数是根据你节点个数来创建的,你节点有几个他就有几个
calico-node-kp8cf                                1/1     Running   18 (60m ago)   11d
calico-node-svqs8                                1/1     Running   18 (60m ago)   11d
calico-node-tbkkf                                1/1     Running   24 (57m ago)   11d
calico-node-wx686                                1/1     Running   18 (57m ago)   11d
calico-node-x2ldl                                1/1     Running   18 (60m ago)   11d
calico-typha-77fc8866f5-2f7k5                    1/1     Running   18 (60m ago)   11d
coredns-567c556887-74dh2                         1/1     Running   18 (57m ago)   11d
coredns-567c556887-xdlw6                         1/1     Running   18 (57m ago)   11d
etcd-k8s-master1                                 1/1     Running   24 (61m ago)   11d
etcd-k8s-master2.guoguo.com                      1/1     Running   18 (57m ago)   11d
etcd-k8s-master3.guoguo.com                      1/1     Running   18 (57m ago)   11d
kube-apiserver-k8s-master1                       1/1     Running   30 (58m ago)   11d
kube-apiserver-k8s-master2.guoguo.com            1/1     Running   17 (57m ago)   11d
kube-apiserver-k8s-master3.guoguo.com            1/1     Running   18 (57m ago)   11d
kube-controller-manager-k8s-master1              1/1     Running   24 (61m ago)   11d
kube-controller-manager-k8s-master2.guoguo.com   1/1     Running   17 (57m ago)   11d
kube-controller-manager-k8s-master3.guoguo.com   1/1     Running   19 (57m ago)   11d
kube-proxy-888wd                                 1/1     Running   24 (61m ago)   11d
#proxy也是根据集群节点个数 来创建的 
kube-proxy-bv22g                                 1/1     Running   18 (57m ago)   11d
kube-proxy-dfflv                                 1/1     Running   18 (57m ago)   11d
kube-proxy-dg8b8                                 1/1     Running   18 (60m ago)   11d
kube-proxy-pnqr8                                 1/1     Running   18 (60m ago)   11d
kube-proxy-whq86                                 1/1     Running   18 (60m ago)   11d
kube-scheduler-k8s-master1                       1/1     Running   26 (61m ago)   11d
kube-scheduler-k8s-master2.guoguo.com            1/1     Running   19 (57m ago)   11d
kube-scheduler-k8s-master3.guoguo.com            1/1     Running   18 (57m ago)   11d
metrics-server-684999f4d6-p6tnj                  1/1     Running   18 (60m ago)   11d

kube-proxy 和calico就是使用的daemonSet

daemonSet不是人为定义副本数量,而是自动通过apiVersion到etcd里面获取你当前集群的节点个数,在etcd里面记录了多少个node节点,daemonSet就会得到这个数量,如果后续添加新的node节点,daemonSet也会创建好对应的pod,如果node下线了daemonSet也会将对应的pod在etcd里面删掉这个就是daemonSet 守护进程集

这个使用的话 通常是做监控使用的

每个集群上要跑一个监控节点,还有网络节点,每个集群上都要运行一个网络容器,不然网络是不通的,还有前面讲的kube-proxy,kube-proxy和APIServer通信的,这个每个节点都得有.

做日志收集使用

daemonSet没有replicas那个值和其他控制器 rc rs deployment 没有什么区别

写一个

[root@k8s-master1 daemonSet]# cat daemonset-nginx-1.yml

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: daemoset-1
spec:
  selector: #这个pod和那些关联,需要用标签来控制pod 所以他也需要标签选择器
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      restartPolicy: Always
      containers:
      - name: image-nginx
        image: images.guoguo.com/apps/nginx:1.22.1
        ports:
        - containerPort: 80
          protocol: TCP

创建

[root@k8s-master1 daemonSet]# kubectl apply -f daemonset-nginx-1.yml

[root@k8s-master1 daemonSet]# kubectl get pods
NAME               READY   STATUS    RESTARTS       AGE
daemoset-1-nwwxj   1/1     Running   0              112s
daemoset-1-qb66h   1/1     Running   0              112s
daemoset-1-sqz8z   1/1     Running   0              112s
#为什么只有三个而不是六个呢,因为当初创建master的时候给master设置了污点,pod创建调度不会去master创建
#我们没有给master定义容忍度,生产上不允许应用程序往master上创建!!!
#除非以后想要监控客户端,那需要在DaemoSet上加上容忍度

回滚

DaemoSet回滚和更新deployment一模一样