k8s学习-DaemonSet和Job

1.1DaemonSet是什么

Deployment部署的副本Pod会分布在各个Node上，每个Node都可能运行好几个副本。DaemonSet的不同之处在于：每个Node上最多只能运行⼀个副本。DaemonSet的典型应用场景有：
（1）在集群的每个节点上运⾏存储Daemon，比如glusterd或ceph。
（2）在每个节点上运行日志收集Daemon，比如flunentd或logstash。
（3）在每个节点上运行监控Daemon，比如PrometheusNodeExporter或collectd。

其实Kubernetes自己就在用DaemonSet运行系统组件。执行如下命令，如下所示。

kubectl get daemonset --namespace=kube-system
kubectl get daemonset --namespace=kube-flannel

[root@k8s-master ~]# kubectl get daemonset --namespace=kube-flannel
NAME              DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
kube-flannel-ds   3         3         3       3            3           <none>          10d
[root@k8s-master ~]# kubectl get daemonset --namespace=kube-flannel
NAME              DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
kube-flannel-ds   3         3         3       3            3           <none>          10d

DaemonSet kube-flannel-ds和kube-proxy分别负责在每个节点上运⾏flannel和kube-proxy组件，在通过查看Pod副本，看看各个节点的分布情况：

kubectl get pod --namespace=kube-system -o wide

[root@k8s-master ~]# kubectl get pod --namespace=kube-system -o wide
NAME                                 READY   STATUS    RESTARTS   AGE   IP                NODE         NOMINATED NODE   READINESS GATES
coredns-7ff77c879f-ltqn4             1/1     Running   2          10d   10.244.0.8        k8s-master   <none>           <none>
coredns-7ff77c879f-xb6qm             1/1     Running   2          10d   10.244.0.9        k8s-master   <none>           <none>
etcd-k8s-master                      1/1     Running   2          10d   192.168.200.128   k8s-master   <none>           <none>
kube-apiserver-k8s-master            1/1     Running   2          10d   192.168.200.128   k8s-master   <none>           <none>
kube-controller-manager-k8s-master   1/1     Running   2          10d   192.168.200.128   k8s-master   <none>           <none>
kube-proxy-dcprp                     1/1     Running   2          10d   192.168.200.128   k8s-master   <none>           <none>
kube-proxy-nklr5                     1/1     Running   2          10d   192.168.200.129   k8s-node1    <none>           <none>
kube-proxy-slslc                     1/1     Running   4          10d   192.168.200.130   k8s-node2    <none>           <none>
kube-scheduler-k8s-master            1/1     Running   2          10d   192.168.200.128   k8s-master   <none>           <none>
metrics-server-7f6b85b597-k6b26      1/1     Running   2          10d   10.244.1.29       k8s-node1    <none>           <none>

1.2 DaemonSet的创建和运行

本小节以Prometheus Node Exporter为例演示用户如何运行自己的DaemonSet。Prometheus 是 流行的系统监控方案 ， Node Exporter 是Prometheus的agent，以Daemon的形式运行在每个被监控节点
上。
vi node-exporter.yaml

apiVersion: apps/v1
kind: DaemonSet
metadata:name: node-exporter-daemonsetnamespace: agent
spec:selector:matchLabels:app: prometheustemplate:metadata:labels:app: prometheusspec:hostNetwork: truecontainers:- name: node-exporterimage: prom/node-exporterimagePullPolicy: IfNotPresentcommand:- /bin/node_exporter- --path.procfs- /host/proc- --path.sysfs- /host/sys-  --collector.filesystem.ignored-mount-points- ^/(sys|proc|dev|host|etc)($|/)volumeMounts:- name: procmountPath: /host/proc- name: sysmountPath: /host/sys- name: rootmountPath: /rootfsvolumes:- name: prochostPath:path: /proc- name: syshostPath:path: /sys- name: roothostPath:path: /

先创建个命名空间

kubectl create namespace agent

通过kubectl创建资源：

kubectl apply -f node-exporter.yaml

然后，通过kubectl查看Pod分布情况：

[root@k8s-master ~]# kubectl get pod --namespace=agent -o wide
NAME                            READY   STATUS    RESTARTS   AGE   IP                NODE        NOMINATED NODE   READINESS GATES
node-exporter-daemonset-5rrq4   1/1     Running   1          44h   192.168.200.129   k8s-node1   <none>           <none>
node-exporter-daemonset-hjkzs   1/1     Running   1          44h   192.168.200.130   k8s-node2   <none>           <none>

2.1 Job

容器按照持续运行的时间可分为两类：服务类容器和工作类容器。
服务类容器通常持续提供服务，需要⼀直运行，比如HTTPServer、Daemon等。工作类容器则是⼀次性任务，比如批处理程序，完成后容器就退出。
Kubernetes的Deployment、ReplicaSet和DaemonSet都用于管理服务类容器；
对于工作类容器，我们使用Job。先看⼀个简单的Job配置文件件myjob.yml。

vi myjob.yml

apiVersion: batch/v1
kind: Job
metadata:name: myjobnamespace: jobs
spec:template:metadata:labels:app: myjobspec:containers:- name: hello-jobimage: busyboximagePullPolicy: IfNotPresentcommand: ["echo", "hello  k8s job!"]restartPolicy: Never

kubectl apply -f myjob.yml

通过查看Job运行情况可以知道，其运行结束就结束了，如下图所示，变成了Completed状态。

kubectl get pod -n jobs

[root@k8s-master ~]# kubectl get pod -n jobs
NAME                              READY   STATUS      RESTARTS   AGE
myjob-krgdh                       0/1     Completed   0          44h

还可以通过查看Log看看这个Job留下的足迹：

kubectl logs myjob-krgdh -n jobs #查看日志
kubectl describe pods/myjob-krgdh --namespace jobs #查看job描述

ng)

[root@k8s-master ~]# kubectl logs myjob-krgdh -n jobs
hello  k8s job!
[root@k8s-master ~]# kubectl describe pods/myjob-krgdh --namespace jobs
Name:         myjob-krgdh
Namespace:    jobs
Priority:     0
Node:         k8s-node2/192.168.200.130
Start Time:   Sat, 27 Jan 2024 16:53:58 +0800
Labels:       app=myjobcontroller-uid=1a312a0b-b6eb-4219-958f-801edc224608job-name=myjob
Annotations:  <none>
Status:       Succeeded
IP:           10.244.2.14
IPs:IP:           10.244.2.14
Controlled By:  Job/myjob
Containers:hello-job:Container ID:  docker://0403ccea759585e45f39431bf3958a169211210ef948a0b7c1d4be1ed411e272Image:         busyboxImage ID:      docker-pullable://busybox@sha256:5acba83a746c7608ed544dc1533b87c737a0b0fb730301639a0179f9344b1678Port:          <none>Host Port:     <none>Command:echohello  k8s job!State:          TerminatedReason:       CompletedExit Code:    0Started:      Sat, 27 Jan 2024 16:53:59 +0800Finished:     Sat, 27 Jan 2024 16:53:59 +0800Ready:          FalseRestart Count:  0Environment:    <none>Mounts:/var/run/secrets/kubernetes.io/serviceaccount from default-token-vz4wm (ro)
Conditions:Type              StatusInitialized       TrueReady             FalseContainersReady   FalsePodScheduled      True
Volumes:default-token-vz4wm:Type:        Secret (a volume populated by a Secret)SecretName:  default-token-vz4wmOptional:    false
QoS Class:       BestEffort
Node-Selectors:  <none>
Tolerations:     node.kubernetes.io/not-ready:NoExecute for 300snode.kubernetes.io/unreachable:NoExecute for 300s
Events:          <none>

2.2 并行Job

如果希望能够同时并行运行多个Pod以提高Job的执行效率，Job提供了一个贴心的配置：parallesim。例如下面的配置，我们将上面的小Job改为并行运行的Pod数量设置为3。
修改 myjob.yml

apiVersion: batch/v1
kind: Job
metadata:name: testjob
spec:parallelism: 3template:metadata:labels:app: testjobspec:containers:- name: hello-jobimage: busyboximagePullPolicy: IfNotPresentcommand: ["echo", "hello  k8s job!"]restartPolicy: Never

kubectl get pod -o wide -n default

可以看出，Job一共启动了3个Pod，都是同时结束的（可以看到三个Pod的AGE都是相同的）。

[root@k8s-master ~]# kubectl get pod -o wide -n default
NAME                                READY   STATUS      RESTARTS   AGE   IP             NODE        NOMINATED NODE   READINESS GATES
testjob-jmlzm                       0/1     Completed   0          44h   10.244.2.18    k8s-node2   <none>           <none>
testjob-r62f4                       0/1     Completed   0          44h   10.244.2.17    k8s-node2   <none>           <none>
testjob-vg7c6                       0/1     Completed   0          44h   10.244.2.16    k8s-node2   <none>           <none>

2.3 定时Job

Linux中有cron程序定时执行任务，Kubernetes的CronJob提供了类似的功能，可以定时执行Job。
vi hello-cron-job.yml

apiVersion: batch/v1beta1
kind: CronJob
metadata:name: hello-cron-jobnamespace: jobs
spec:schedule: "*/1 * * * *"jobTemplate:spec:template:spec:containers:- name: hello-cron-jobimage: busyboximagePullPolicy: IfNotPresentcommand: ["echo", "hello edison's k8s cron job!"]restartPolicy: OnFailure

上面加粗的配置是CronJob的独有配置，需要注意的是schedule，它的格式和Linux Cron一样，这里的"*/1 * * * *"代表每一分钟启动执行一次。对于CronJob，它需要的是jobTemplate来定义Job的模板。

kubectl get cronjob -n jobs

[root@k8s-master ~]# kubectl get cronjob -n jobs
NAME             SCHEDULE      SUSPEND   ACTIVE   LAST SCHEDULE   AGE
hello-cron-job   */1 * * * *   False     0        43h             43h

查看job

可以看到，在过去的三分钟里，每一分钟都启动了一个Pod，符合预期。

kubectl get jobs -n jobs

[root@k8s-master ~]# kubectl get jobs -n jobs
NAME                        COMPLETIONS   DURATION   AGE
hello-cron-job-1706350440   1/1           1s         43h
hello-cron-job-1706350500   1/1           1s         43h
hello-cron-job-1706350560   1/1           1s         43h