Kubernetes 副本機制篇

Kubernetes 副本機制篇

前言

在上文Kubernetes Pod操作篇介紹了kubernetes的核心組件Pod,本文繼續介紹kubernetes的副本機制,正是因為副本機制你的部署能自動保待運行,並且保持健康,無須任何手動干預。

探針

kubernetes可以通過存活探針(liveness probe)檢查容器是否還在運行。可以為pod中的每個容器單獨指定存活探針;如果探測失敗,kubernetes將定期執行探針並重新啟動容器;
kubernetes有以下三種探測容器的機制:

  • HTTP GET探針對容器的IP地址執行HTTP GET請求;
  • TCP套接字探針嘗試與容器指定端口建立TCP連接;
  • Exec探針在容器內執行任意命令,並檢查命令的退出狀態碼。

1.準備鏡像

1.1 準備App.js

為了測試探針的作用,需要準備新的鏡像;在之前的服務中稍作改動,在第五個請求之後,給每個請求返回HTTP狀態碼500(Internal Server Error),app.js做如下改動:

<code>const http = require('http');
const os = require('os');
console.log("kubia server is starting...");
var requestCount = 0;
var handler = function(request,response){
    console.log("Received request from " + request.connection.remoteAddress);
    requestCount++;
    if (requestCount > 5) {
      response.writeHead(500);
      response.end("I'm not well. Please restart me!");
      return;
    }
    response.writeHead(200);
    response.end("You've hit " + os.hostname()+"\n");
};
var www = http.createServer(handler);
www.listen(8080);/<code>

requestCount記錄請求的次數,大於5次直接返回500狀態碼,這樣探針可以捕獲狀態碼進行服務器重啟;

1.2 構建鏡像

<code>[root@localhost unhealthy]# docker build -t kubia-unhealthy .
Sending build context to Docker daemon  3.584kB
Step 1/3 : FROM node:7
 ---> d9aed20b68a4
Step 2/3 : ADD app.js /app.js
 ---> e9e1b44f8f54
Step 3/3 : ENTRYPOINT ["node","app.js"]
 ---> Running in f58d6ff6bea3
Removing intermediate container f58d6ff6bea3
 ---> d36c6390ec66
Successfully built d36c6390ec66
Successfully tagged kubia-unhealthy:latest/<code>

通過docker build構建kubia-unhealthy鏡像

1.3 推送鏡像

<code>[root@localhost unhealthy]# docker tag kubia-unhealthy ksfzhaohui/kubia-unhealthy
[root@localhost unhealthy]# docker login
Authenticating with existing credentials...
WARNING! Your password will be stored unencrypted in /root/.docker/config.json.
Configure a credential helper to remove this warning. See
https://docs.docker.com/engine/reference/commandline/login/#credentials-store

Login Succeeded
[root@localhost unhealthy]# docker push ksfzhaohui/kubia-unhealthy
The push refers to repository [docker.io/ksfzhaohui/kubia-unhealthy]
40d9e222a827: Pushed 
......
latest: digest: sha256:5fb3ebeda7f98818bc07b2b1e3245d6a21014a41153108c4dcf52f2947a4dfd4 size: 2213/<code>

首先給鏡像附加標籤,然後登錄docker hub,最後推送到docker hub:

Kubernetes 副本機制篇

2.探針實戰

2.1 Http探針YAML文件

創建YAML描述文件,指定了一個Http Get存活探針,告訴Kubernetes定期在端口路徑下執行Http Get請求,以確定容器是否健康;

<code>apiVersion: v1
kind: Pod
metadata: 
   name: kubia-liveness
spec: 
   containers: 
   - image: ksfzhaohui/kubia-unhealthy
     name: kubia
     livenessProbe: 
        httpGet: 
           path: /
           port: 8080/<code>

2.2 創建Pod

<code>[d:\k8s]$ kubectl create -f kubia-liveness-probe.yaml
pod/kubia-liveness created

[d:\k8s]$ kubectl get pods
NAME             READY   STATUS              RESTARTS   AGE
kubia-liveness   0/1     ContainerCreating   0          3s/<code>

創建名稱為kubia-liveness的Pod,查看的RESTARTS為0,隔一段時間再次觀察:

<code>[d:\k8s]$ kubectl get pods
NAME             READY   STATUS    RESTARTS   AGE
kubia-liveness   1/1     Running   2          4m/<code>

觀察可以發現此時的RESTARTS=2,表示重啟了2次,因為每次探測都會發送http請求,而服務在接收5次請求之後會返回500狀態碼,Kubernetes探測之後就會重啟容器;

2.3 Pod探針描述

<code>[d:\k8s]$ kubectl describe po kubia-liveness
Name:         kubia-liveness
......
    State:          Running
      Started:      Mon, 23 Dec 2019 15:42:45 +0800
    Last State:     Terminated
      Reason:       Error
      Exit Code:    137
      Started:      Mon, 23 Dec 2019 15:41:15 +0800
      Finished:     Mon, 23 Dec 2019 15:42:42 +0800
    Ready:          True
    Restart Count:  2
    Liveness:       http-get http://:8080/ delay=0s timeout=1s period=10s #success=1 #failure=3
......
Events:
  Type     Reason     Age                    From               Message
  ----     ------     ----                   ----               -------
 ......
  Warning  Unhealthy  85s (x9 over 5m5s)     kubelet, minikube  Liveness probe failed: HTTP probe failed with statuscode: 500
  Normal   Killing    85s (x3 over 4m45s)    kubelet, minikube  Container kubia failed liveness probe, will be restarted
....../<code>

State:當前狀態是運行中;
Last State:最後的狀態是終止,原因是出現了錯誤,退出代碼為137有特殊的含義:表示該進程由外部信號終止,數字137是兩個數字的總和:128+x, 其中x是終止進程的信號編號,這裡x=9是SIGKILL的信號編號,意味著這個進程被強行終止;


Restart Count:重啟的次數;
Liveness:存活探針的附加信息,delay(延遲)、timeout(超時)、period(週期);大致意思就是開始探測延遲為0秒,探測超時時間為1秒,每隔10秒檢測一次,探測連續失敗三次重啟容器;定義探針時可以自定義這些參數,比如initialDelaySeconds設置初始延遲等;
Events:列出了發生的事件,比如探測到失敗,殺進程,重啟容器等;

3.探針總結

首先生產環境運行的pod一定要配置探針;其次探針一定要檢查程序的內部,不受外部因數影響比如外部服務,數據庫等;最後就是探針應該足夠輕量。
以上方式創建的pod,kubernetes在使用探針發現服務不可能就會重啟服務,這項任務由承載pod的節點上的Kubelet執行,在主服務器上運行的Kubernetes Control Plane組件不會參與此過程;但如果節點本身崩潰,由於Kubelet本身運行在節點上,所以如果節點異常終止,它將無法執行任何操作,這時候就需要ReplicationController或類似機制管理pod。

ReplicationController

ReplicationController是一種kubernetes資源,可確保它的pod始終保持運行狀態;如果pod因任何原因消失(包括節點崩潰),則ReplicationController會重新創建Pod;
ReplicationController會持續監控正在運行的pod列表,是確保pod的數量始終與其標籤選擇器匹配,一個ReplicationController有三個主要部分:

  • label selector(標籤選擇器),用於確定ReplicationController作用域中有哪些pod;
  • replica count(副本個數),指定應運行的pod數量;
  • pod template(pod模板),用於創建新的pod副本。

以上三個屬性可以隨時修改,但是隻有副本個數修改對當前pod會有影響,比如當前副本數量減少了,那當前pod有可能會被刪除;ReplicationController提供的好處:

  • 確保一個pod(或多個pod副本)持續運行,失敗重啟新pod;
  • 集群節點發生故障時,它將為故障節點上運行的所有pod創建副本;
  • 輕鬆實現pod的水平伸縮。

1.創建ReplicationController

<code>apiVersion: v1
kind: ReplicationController
metadata: 
   name: kubia
spec: 
   replicas: 3
   selector: 
      app: kubia
   template:
      metadata: 
         labels:
            app: kubia
      spec: 
         containers: 
         - name: kubia
           image: ksfzhaohui/kubia
           ports: 
           - containerPort: 8080/<code>

指定了類型為ReplicationController,名稱為kubia;replicas設置副本為3,selector為標籤選擇器,template為pod創建的模版,三個要素都指定了,執行創建命令:

<code>[d:\k8s]$ kubectl create -f kubia-rc.yaml
replicationcontroller/kubia created

[d:\k8s]$ kubectl get pods
NAME          READY   STATUS    RESTARTS   AGE
kubia-dssvz   1/1     Running   0          73s
kubia-krlcr   1/1     Running   0          73s
kubia-tg29c   1/1     Running   0          73s/<code>

創建完之後等一會執行獲取pod列表可以發現創建了三個容器,刪除其中一個,再次觀察:

<code>[d:\k8s]$ kubectl delete pod kubia-dssvz
pod "kubia-dssvz" deleted

[d:\k8s]$ kubectl get pods
NAME          READY   STATUS        RESTARTS   AGE
kubia-dssvz   1/1     Terminating   0          2m2s
kubia-krlcr   1/1     Running       0          2m2s
kubia-mgz64   1/1     Running       0          11s
kubia-tg29c   1/1     Running       0          2m2s/<code>

被刪除的pod結束中,新的pod已經啟動,獲取有關ReplicationController的信息:

<code>[d:\k8s]$ kubectl get rc
NAME    DESIRED   CURRENT   READY   AGE
kubia   3         3         3       4m20s/<code>

期望3個副本,當前3個副本,準備好的也是3個,更詳細的可以使用describe命令:

<code>[d:\k8s]$ kubectl describe rc kubia
Name:         kubia
Namespace:    default
Selector:     app=kubia
Labels:       app=kubia
Annotations:  
Replicas:     3 current / 3 desired
Pods Status:  3 Running / 0 Waiting / 0 Succeeded / 0 Failed
Pod Template:
......
Events:
  Type    Reason            Age    From                    Message
  ----    ------            ----   ----                    -------
  Normal  SuccessfulCreate  5m20s  replication-controller  Created pod: kubia-dssvz
  Normal  SuccessfulCreate  5m20s  replication-controller  Created pod: kubia-tg29c
  Normal  SuccessfulCreate  5m20s  replication-controller  Created pod: kubia-krlcr
  Normal  SuccessfulCreate  3m29s  replication-controller  Created pod: kubia-mgz64
  Normal  SuccessfulCreate  75s    replication-controller  Created pod: kubia-vwnmf/<code>

Replicas顯示副本期望數和當前數,Pods Status顯示每種狀態下的副本數,最後的Events為發生的事件,測試一共刪除2個pod,可以看到一個創建了5個pod;

注:因為使用的是Minikube,只有一個節點同時充當主節點和工作節點,節點故障無法模擬。

2.修改標籤

通過更改pod的標籤,可以將它從ReplicationController的作用域中添加或刪除:

<code>[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE   LABELS
kubia-mgz64   1/1     Running   0          27m   app=kubia
kubia-tg29c   1/1     Running   0          28m   app=kubia
kubia-vwnmf   1/1     Running   0          24m   app=kubia

[d:\k8s]$ kubectl label pod kubia-mgz64 app=foo --overwrite
pod/kubia-mgz64 labeled

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS              RESTARTS   AGE   LABELS
kubia-4dzw8   0/1     ContainerCreating   0          2s    app=kubia
kubia-mgz64   1/1     Running             0          27m   app=foo
kubia-tg29c   1/1     Running             0          29m   app=kubia
kubia-vwnmf   1/1     Running             0          25m   app=kubia/<code>

可以發現初始創建的是三個Pod標籤都是app=kubia,當把kubia-mgz64的標籤設置為foo之後就脫離了當前ReplicationController的控制,這樣ReplicationController控制的副本就變成了2個,所以會里面重新創建一個Pod;脫離控制的Pod還是照常運行,除非我們手動刪除;

<code>[d:\k8s]$ kubectl delete pod kubia-mgz64
pod "kubia-mgz64" deleted

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE   LABELS
kubia-4dzw8   1/1     Running   0          20h   app=kubia
kubia-tg29c   1/1     Running   0          21h   app=kubia
kubia-vwnmf   1/1     Running   0          21h   app=kubia/<code>

3.修改Pod模版

ReplicationController的pod模板可以隨時修改:

<code>[d:\k8s]$ kubectl edit rc kubia
......
replicationcontroller/kubia edited/<code>

使用如上命令即可,會彈出文本編輯器,修改Pod模版標籤,如下所示:

<code>  template:
    metadata:
      creationTimestamp: null
      labels:
        app: kubia
        type: special/<code>

添加新的標籤type:special,保存退出即可;修改Pod模版之後並不影響現有的pod,只會影響重新創建的pod:

<code>[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE   LABELS
kubia-4dzw8   1/1     Running   0          21h   app=kubia
kubia-tg29c   1/1     Running   0          21h   app=kubia
kubia-vwnmf   1/1     Running   0          21h   app=kubia

[d:\k8s]$ kubectl delete pod kubia-4dzw8
pod "kubia-4dzw8" deleted

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE     LABELS
kubia-6qrxj   1/1     Running   0          2m12s   app=kubia,type=special
kubia-tg29c   1/1     Running   0          21h     app=kubia
kubia-vwnmf   1/1     Running   0          21h     app=kubia/<code>

刪除一個pod,重新創建的pod有了新的標籤;

4.水平縮放pod

通過文本編輯器來修改副本數,修改spec.replicas為5

<code>[d:\k8s]$ kubectl edit rc kubia
replicationcontroller/kubia edited

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS              RESTARTS   AGE     LABELS
kubia-6qrxj   1/1     Running             0          9m49s   app=kubia,type=special
kubia-9crmf   0/1     ContainerCreating   0          4s      app=kubia,type=special
kubia-qpwbl   0/1     ContainerCreating   0          4s      app=kubia,type=special
kubia-tg29c   1/1     Running             0          21h     app=kubia
kubia-vwnmf   1/1     Running             0          21h     app=kubia/<code>

可以發現自動創建了2個Pod,達到副本數5;通過kubectl scale重新修改為3:

<code>[d:\k8s]$ kubectl scale rc kubia --replicas=3
replicationcontroller/kubia scaled

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE   LABELS
kubia-6qrxj   1/1     Running   0          15m   app=kubia,type=special
kubia-tg29c   1/1     Running   0          22h   app=kubia
kubia-vwnmf   1/1     Running   0          21h   app=kubia/<code>

5.刪除ReplicationController

通過kubectl delete刪除ReplicationController時默認會刪除pod,但是也可以指定不刪除:

<code>[d:\k8s]$ kubectl delete rc kubia --cascade=false
replicationcontroller "kubia" deleted

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE    LABELS
kubia-6qrxj   1/1     Running   0          103m   app=kubia,type=special
kubia-tg29c   1/1     Running   0          23h    app=kubia
kubia-vwnmf   1/1     Running   0          23h    app=kubia

[d:\k8s]$ kubectl get rc kubia
Error from server (NotFound): replicationcontrollers "kubia" not found/<code>

--cascade=false可以不刪除pod,只刪除ReplicationController

ReplicaSet

ReplicaSet是新一代ReplicationController,將完全替代ReplicationController;ReplicaSet的行為與ReplicationController完全相同,但pod選擇器的表達能力更強;

1.創建ReplicaSet

<code>apiVersion: apps/v1
kind: ReplicaSet
metadata: 
   name: kubia
spec: 
   replicas: 3
   selector: 
      matchLabels: 
         app: kubia      
   template:
      metadata: 
         labels:
            app: kubia
      spec: 
         containers: 
         - name: kubia
           image: ksfzhaohui/kubia/<code>

apiVersion指定為apps/v1:apps表示API組,v1表示實際的API版本;如果是在核心的API組中,API是可以不用指定的,比如之前的ReplicationController只需要指定v1;
其他定義基本和ReplicationController類似,除了在selector下使用了matchLabels選擇器;

<code>[d:\k8s]$ kubectl create -f kubia-replicaset.yaml
replicaset.apps/kubia created

[d:\k8s]$ kubectl get pods --show-labels
NAME          READY   STATUS    RESTARTS   AGE    LABELS
kubia-6qrxj   1/1     Running   0          150m   app=kubia,type=special
kubia-tg29c   1/1     Running   0          24h    app=kubia
kubia-vwnmf   1/1     Running   0          24h    app=kubia

[d:\k8s]$ kubectl get rs
NAME    DESIRED   CURRENT   READY   AGE
kubia   3         3         3       49s/<code>

創建完ReplicaSet之後,重新接管了原來的3個pod;更詳細的可以使用describe命令:

<code>[d:\k8s]$ kubectl describe rs
Name:         kubia
Namespace:    default
Selector:     app=kubia
Labels:       
Annotations:  
Replicas:     3 current / 3 desired
Pods Status:  3 Running / 0 Waiting / 0 Succeeded / 0 Failed
Pod Template:
  Labels:  app=kubia
  Containers:
   kubia:
    Image:        ksfzhaohui/kubia
    Port:         
    Host Port:    
    Environment:  
    Mounts:       
  Volumes:        
Events:           /<code>

可以看到Events事件列表為空,當前的3個pod都是接管的原來已經創建的pod;

2.ReplicaSet標籤選擇器

ReplicaSet相對於ReplicationController的主要改進是它更具表達力的標籤選擇器;

<code>   selector: 
      matchExpressions:
       - key: app
         operator: In
         values: 
            - kubia/<code>

ReplicaSet除了可以使用matchLabels,還可以使用功能更強大的matchExpressions;每個表達式都必須包含一個key、一個operator(運算符)、可能還有一個values的列表,運算符可以有:

  • In:Label的值必須與其中一個指定的values匹配;
  • Notln:Label的值與任何指定的values不匹配;
  • Exists:pod必須包含一個指定名稱的標籤,使用此運算符時,不應指定values字段;
  • DoesNotExist:pod不得包含有指定名稱的標籤,不應指定values字段;

3.刪除ReplicaSet

<code>[d:\k8s]$ kubectl delete rs kubia
replicaset.apps "kubia" deleted

[d:\k8s]$ kubectl get pods --show-labels
No resources found in default namespace./<code>

刪除ReplicaSet的同時會刪除其管理的pod;

DaemonSet

Replicationcontroller和ReplicaSet都用於在kubernetes集群上運行部署特定數量的pod;而DaemonSet可以在所有集群節點上運行一個pod,比如希望在每個節點上運行日誌收集器和資源監控器;當然也可以通過節點選擇器控制只有哪些節點運行pod;

1.創建DaemonSet

<code>apiVersion: apps/v1
kind: DaemonSet
metadata: 
   name: ssd-monitor
spec: 
   selector: 
      matchLabels: 
         app: ssd-monitor
   template:
      metadata: 
         labels:
           app: ssd-monitor
      spec: 
         nodeSelector: 
           disk: ssd
         containers: 
         - name: main
           image: ksfzhaohui/kubia/<code>

準備如上創建DaemonSet的YAML文件,以上屬性基本和ReplicaSet類似,除了nodeSelector也就是節點選擇器,指定了選擇disk=ssd標籤;
的節點標籤;

<code>[d:\k8s]$ kubectl create -f ssd-monitor-daemonset.yaml
daemonset.apps/ssd-monitor created

[d:\k8s]$ kubectl get ds
NAME          DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
ssd-monitor   0         0         0       0            0           disk=ssd        24s

[d:\k8s]$ kubectl get pods --show-labels
No resources found in default namespace./<code>

創建完之後,並沒有給當前節點創建pod,因為當前節點沒有指定disk=ssd標籤;

<code>[d:\k8s]$ kubectl get node
NAME       STATUS   ROLES    AGE   VERSION
minikube   Ready    master   8d    v1.17.0

[d:\k8s]$ kubectl label node minikube disk=ssd
node/minikube labeled

[d:\k8s]$ kubectl get node --show-labels
NAME       STATUS   ROLES    AGE   VERSION   LABELS
minikube   Ready    master   8d    v1.17.0   beta.kubernetes.io/arch=amd64,beta.kubernetes.io/os=linux,disk=ssd,gpu=true,kubernetes.io/arch=amd64,kubernetes.io/hostname=minikube,kubernetes.io/os=linux,node-role.kubernetes.io/master=

[d:\k8s]$ kubectl get pods --show-labels
NAME                READY   STATUS    RESTARTS   AGE   LABELS
ssd-monitor-84hxd   1/1     Running   0          31s   app=ssd-monitor,controller-revision-hash=5dc77f567d,pod-template-generation=1/<code>

首先獲取當前節點名稱為minikube,然後設置標籤disk=ssd,這時候會自動在當前節點創建一個pod,因為在minikube中只有一個節點不好在多個節點上模擬;

2.刪除pod和DaemonSet

<code>[d:\k8s]$ kubectl label node minikube disk=hdd --overwrite
node/minikube labeled

[d:\k8s]$ kubectl get pods --show-labels
No resources found in default namespace./<code>

修改節點minkube的標籤,可以發現節點上的pod會自動刪除,因為不滿足節點選擇器;

<code>[d:\k8s]$ kubectl delete ds ssd-monitor
daemonset.apps "ssd-monitor" deleted

[d:\k8s]$ kubectl get ds
No resources found in default namespace./<code>

刪除DaemonSet也會一起刪除這些pod;

Job

ReplicationController、ReplicaSet和DaemonSet會持續運行任務,永遠達不到完成態,這些pod中的進程在退出時會重新啟動;kubernetes通過Job資源允許你運行一種pod, 該pod在內部進程成功結束時,不重啟容器,一旦任務完成,pod就被認為處千完成狀態;
在發生節點故障時,該節點上由Job管理的pod,重新安排到其他節點;如果進程本身異常退出,可以將Job配置為重新啟動容器;

1.創建Job

在創建Job前先準備一個構建在busybox的鏡像,該容器將調用sleep 命令兩分鐘:

<code>FROM busybox
ENTRYPOINT echo "$(date) Batch job starting"; sleep 120; echo "$(date) Finished succesfully"/<code>

此鏡像已經推送到docker hub:

Kubernetes 副本機制篇

<code>apiVersion: batch/v1
kind: Job
metadata: 
   name: batch-job
spec: 
   template:
      metadata: 
         labels:
           app: batch-job
      spec: 
         restartPolicy: OnFailure
         containers: 
         - name: main
           image: ksfzhaohui/batch-job/<code>

Job屬於batch API組,其中重要的屬性是restartPolicy默認為Always表示無限期運行,其他選項還有OnFailure或Never,表示進程失敗重啟和不重啟;

<code>[d:\k8s]$ kubectl create -f exporter.yaml
job.batch/batch-job created

[d:\k8s]$ kubectl get job
NAME        COMPLETIONS   DURATION   AGE
batch-job   0/1           7s         8s

[d:\k8s]$ kubectl get pod
NAME              READY   STATUS    RESTARTS   AGE
batch-job-7sw68   1/1     Running   0          25s/<code>

創建Job,會自動創建一個pod,pod中的進程運行2分鐘後會結束:

<code>[d:\k8s]$ kubectl get pod
NAME              READY   STATUS      RESTARTS   AGE
batch-job-7sw68   0/1     Completed   0          3m1s

[d:\k8s]$ kubectl get job
NAME        COMPLETIONS   DURATION   AGE
batch-job   1/1           2m11s      3m12s/<code>

可以發現pod狀態為Completed,同樣job的COMPLETIONS同樣為完成;

2.Job中運行多個pod實例

作業可以配置為創建多個pod實例,並以並行或串行方式運行它們;可以通過設置completions和parallelism屬性來完成;

2.1 順序運行Job pod

<code>apiVersion: batch/v1
kind: Job
metadata: 
   name: multi-completion-batch-job
spec: 
   completions: 3
   template:
      metadata: 
         labels:
           app: multi-completion-batch-job
      spec: 
         restartPolicy: OnFailure
         containers: 
         - name: main
           image: ksfzhaohui/batch-job/<code>

completions設置為3,一個一個的運行3個pod,所有完成整個job完成;

<code>[d:\k8s]$ kubectl get pod
NAME                               READY   STATUS      RESTARTS   AGE
multi-completion-batch-job-h75j8   0/1     Completed   0          2m19s
multi-completion-batch-job-wdhnj   1/1     Running     0          15s

[d:\k8s]$ kubectl get job
NAME                         COMPLETIONS   DURATION   AGE
multi-completion-batch-job   1/3           2m28s      2m28s/<code>

可以看到完成一個pod之後會啟動第二pod,所有都運行完之後如下所示:

<code>[d:\k8s]$ kubectl get pod
NAME                               READY   STATUS      RESTARTS   AGE
multi-completion-batch-job-4vjff   0/1     Completed   0          2m7s
multi-completion-batch-job-h75j8   0/1     Completed   0          6m16s
multi-completion-batch-job-wdhnj   0/1     Completed   0          4m12s

[d:\k8s]$ kubectl get job
NAME                         COMPLETIONS   DURATION   AGE
multi-completion-batch-job   3/3           6m13s      6m18s/<code>

2.2 並行運行Job pod

<code>apiVersion: batch/v1
kind: Job
metadata: 
   name: multi-completion-parallel-batch-job
spec: 
   completions: 3
   parallelism: 2
   template:
      metadata: 
         labels:
           app: multi-completion-parallel-batch-job
      spec: 
         restartPolicy: OnFailure
         containers: 
         - name: main
           image: ksfzhaohui/batch-job/<code>

同時設置了completions和parallelism,表示job可以同時運行兩個pod,其中任何一個執行完成可以運行第三個pod:

<code>[d:\k8s]$ kubectl create -f multi-completion-parallel-batch-job.yaml
job.batch/multi-completion-parallel-batch-job created

[d:\k8s]$ kubectl get pod
NAME                                        READY   STATUS              RESTARTS   AGE
multi-completion-parallel-batch-job-f7wn8   0/1     ContainerCreating   0          3s
multi-completion-parallel-batch-job-h9s29   0/1     ContainerCreating   0          3s/<code>

2.3 限制Job pod完成任務的時間

在pod配置中設置activeDeadlineSeconds屬性,可以限制pod的時間;如果pod運行時間超過此時間,系統將嘗試終止pod, 並將Job標記為失敗;

<code>apiVersion: batch/v1
kind: Job
metadata:
  name: time-limited-batch-job
spec:
  activeDeadlineSeconds: 30
  template:
    metadata:
      labels:
        app: time-limited-batch-job
    spec:
      restartPolicy: OnFailure
      containers:
      - name: main
        image: ksfzhaohui/batch-job/<code>

指定activeDeadlineSeconds為30秒,超過30秒自動失敗;

<code>[d:\k8s]$ kubectl create -f time-limited-batch-job.yaml
job.batch/time-limited-batch-job created

[d:\k8s]$ kubectl get job
NAME                     COMPLETIONS   DURATION   AGE
time-limited-batch-job   0/1           3s         3s

[d:\k8s]$ kubectl get pod
NAME                           READY   STATUS    RESTARTS   AGE
time-limited-batch-job-jgmm6   1/1     Running   0          29s

[d:\k8s]$ kubectl get pod
NAME                           READY   STATUS        RESTARTS   AGE
time-limited-batch-job-jgmm6   1/1     Terminating   0          30s

[d:\k8s]$ kubectl get pod
No resources found in default namespace.

[d:\k8s]$ kubectl get job
NAME                     COMPLETIONS   DURATION   AGE
time-limited-batch-job   0/1           101s       101s/<code>

可以觀察AGE標籤下面的時間表示已經運行的時間,30秒之後pod狀態變成Terminating;

2.4 Job定期運行

job也支持定期執行,有點像quartz,也支持類似的quartz表達式:

<code>apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: corn-batch-job
spec:
  schedule: "0-59 * * * *"
  jobTemplate:
    spec:
      template:
        metadata:
          labels:
            app: corn-batch-job
        spec:
          restartPolicy: OnFailure
          containers:
          - name: main
            image: ksfzhaohui/batch-job/<code>

指定schedule用來表示表達式分別是:分鐘,小時,每個月中的第幾天,月,星期幾;以上配置表示每分鐘運行一個job;

<code>[d:\k8s]$ kubectl create -f cronjob.yaml
cronjob.batch/corn-batch-job created

[d:\k8s]$ kubectl get pod
NAME                              READY   STATUS              RESTARTS   AGE
corn-batch-job-1577263560-w2fq2   0/1     Completed           0          3m3s
corn-batch-job-1577263620-92pc7   1/1     Running             0          2m2s
corn-batch-job-1577263680-tmr8p   1/1     Running             0          62s
corn-batch-job-1577263740-jmzqk   0/1     ContainerCreating   0          2s

[d:\k8s]$ kubectl get job
NAME                        COMPLETIONS   DURATION   AGE
corn-batch-job-1577263560   1/1           2m5s       3m48s
corn-batch-job-1577263620   1/1           2m4s       2m47s
corn-batch-job-1577263680   0/1           107s       107s
corn-batch-job-1577263740   0/1           47s        47s/<code>

每個一分鐘就運行一個job,可以刪除CronJob

<code>[d:\k8s]$ kubectl delete CronJob corn-batch-job
cronjob.batch "corn-batch-job" deleted/<code>

總結

本文繼續在閱讀Kubernetes in Action過程中,實際操作的筆記;主要介紹了相關的副本機制探針,ReplicationController,ReplicaSet,DaemonSet以及Job相關知識點。

參考

Kubernetes in Action

博客地址

https://github.com/ksfzhaohui/blog


分享到:


相關文章: