本篇是Docker&Kubenetes系列的第四篇,在前面的篇幅中,我们向Kubenetes中部署了单个应用实例。如果单个节点故障了,那么服务就不可用了,这在实际环境中是不能接受的。在实际的正式环境中,我们不仅要避免单点,还要根据负载变化动态的调整节点数量。为了实现这个目的,我们可以借助于Kubenetes的Deployment,Deployment可以创建指定数量的Pod,并有自愈能力,还能完成升级更新及回滚操作。
向Kubenetes中部署一个3节点的集群
和之前介绍的一样部署集群也由过yml配置文件开始的,本例我们依然使用前面篇幅中创建的docker镜像来部署集群。Deployment配置文件内容如下,这个配置文件是从Kubenetes官网复制的,手写yml文件太容易出错了,yml格式挺烦的。
apiVersion: apps/v1
kind: Deployment
metadata:
name: my-first-demo-dep
labels:
app: my-first-demo
spec:
replicas: 3
selector:
matchLabels:
app: my-first-demo
template:
metadata:
labels:
app: my-first-demo
spec:
containers:
- name: my-first-demo
image: hellodm/my-first-demo:v1.0
ports:
- containerPort: 80
配置文件中,replicas=3
表示要启动3个副本,使用的镜像是hellodm/my-first-demo:v1.0
。配置完了之后,我们来启动集群,命令如下:
$ kubectl create -f dep.yml
deployment.apps/my-first-demo-dep created
我们来看查看副本集情况:
$ kubectl get rs
NAME DESIRED CURRENT READY AGE
my-first-demo-dep-5647fd55f 3 3 3 3m15s
在查看一下pod:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
my-first-demo-dep-5647fd55f-4wzxs 1/1 Running 0 22s
my-first-demo-dep-5647fd55f-c9lwx 1/1 Running 0 22s
my-first-demo-dep-5647fd55f-nnwt6 1/1 Running 0 22s
暴露集群,使得我们可以在访问集群内部的应用。
$ kubectl apply -f svc.yml
service/my-first-demo-svc created
访问:
$ curl 'http://localhost:30000/'
<h1>Hello world! <h1>
OK,成功了。
Kubenetes 扩容演示
上面我们部署了3个节点的集群,假设现在3个节点已经有点吃力了,我们准备扩展到4个节点。编辑上面的Deployment文件,修改的地方如下:
spec:
replicas: 4
其他的不变,只是replicas
为4了。重新启动一下看看:
$ kubectl apply -f dep.yml --record=true
deployment.apps/my-first-demo-dep configured
看下副本集,如下,变成4个了。
$ kubectl get rs
NAME DESIRED CURRENT READY AGE
my-first-demo-dep-5647fd55f 4 4 4 10h
检查Kubenetes集群的自愈性
我们来删掉2个Pod,模拟宕机情况,执行kubectl delete pod
命令,如下:
$ kubectl delete pod my-first-demo-dep-5647fd55f-4wzxs my-first-demo-dep-5647fd55f-c9lwx
立马查看Pod是否恢复:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
my-first-demo-dep-5647fd55f-bxpn7 1/1 Running 0 25s
my-first-demo-dep-5647fd55f-nnwt6 1/1 Running 0 5m8s
my-first-demo-dep-5647fd55f-vwxgz 1/1 Running 0 25s
可以看到还是3台,从NAME和AGE可以看出,多了两个新的。OK,至此,我们演示了扩容和自愈性。
Kubenetes滚动升级
为了演示滚动升级,我们先制作v2.0版本的镜像,修改index的内容为:
<h1>Hello worldls , this is Version 2.</h1>
这里我故意拼错了一个单词,别急,后面有妙用。下面我们来
制作v2.0的镜像:
$ docker build . -t my-frist-demo:v2.0
Sending build context to Docker daemon 6.144kB
Step 1/2 : FROM nginx
---> 602e111c06b6
Step 2/2 : COPY index.html /usr/share/nginx/html
---> 7484909c0df2
Successfully built 7484909c0df2
Successfully tagged my-frist-demo:v2.0
发布镜像:
$ docker tag my-frist-demo:v2.0 hellodm/my-first-demo:v2.0
$ docker push hellodm/my-first-demo:v2.0
The push refers to repository [docker.io/hellodm/my-first-demo]
a3e37d09f192: Preparing
b3003aac411c: Preparing
216cf33c0a28: Preparing
c2adabaecedb: Layer already exists
denied: requested access to the resource is denied
修改deployment的配置文件,改用v2.0的镜像,如下:
apiVersion: apps/v1
kind: Deployment
metadata:
name: my-first-demo-dep
labels:
app: my-first-demo
spec:
replicas: 3
selector:
matchLabels:
app: my-first-demo
template:
metadata:
labels:
app: my-first-demo
spec:
containers:
- name: my-first-demo
image: hellodm/my-first-demo:v2.0
ports:
- containerPort: 80
执行发布,这次附带了一个参数--record=true
让 Kubernetes 把这行命令记到发布历史中方便后面查看。
$ kubectl apply -f dep.yml --record=true
deployment.apps/my-first-demo-dep configured
赶紧查看一下pod状态如下,可以发现已经在滚动发布了,ContainerCreating
状态的表示新版本的容器正在启动,Running
的是新版本的已经运行起来了,Terminating
的这个是老版本正在停止。
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
my-first-demo-dep-54596847d8-6m4n9 0/1 ContainerCreating 0 2s
my-first-demo-dep-54596847d8-jrm8g 1/1 Running 0 4s
my-first-demo-dep-5647fd55f-cbcrz 1/1 Running 0 56s
my-first-demo-dep-5647fd55f-ll7tt 0/1 Terminating 0 56s
my-first-demo-dep-5647fd55f-s8d6l 1/1 Running 0 56s
再赶快查看一下滚动发布的状态:
$ kubectl rollout status deployment my-first-demo-dep
Waiting for rollout to finish: 1 old replicas are pending termination...
deployment "my-first-demo-dep" successfully rolled out
如上命令所表示,随着最后一个old replicas
终止,新版本的deployment成功发布了successfully rolled out
。
简直是太激动了,验证一下是不是3个pod,如下,没问题,3个Running状态。
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
my-first-demo-dep-54596847d8-6m4n9 1/1 Running 0 36s
my-first-demo-dep-54596847d8-jrm8g 1/1 Running 0 38s
my-first-demo-dep-54596847d8-nqgmn 1/1 Running 0 34s
访问一下看看:
$ curl 'http://localhost:30000/'
<h1>Hello worldls , this is Version 2.</h1>
已经是新版本了,可惜啊,新版本有问题啊!单词拼错了,world
写成了worldls
,我们来赶紧回滚一下。
Kubenetes应用集群回滚
上面我们使用滚动发布,将网站升级为version2了。但是有错误,我们来查看一下发布历史:
$ kubectl rollout history deployment my-first-demo-dep
deployment.extensions/my-first-demo-dep
REVISION CHANGE-CAUSE
1 kubectl apply --filename=dep.yml --record=true
2 kubectl apply --filename=dep.yml --record=true
可以看到,有两个版本,我们回滚到版本1,执行如下命令:
$ kubectl rollout undo deployment my-first-demo-dep --to-revision=1
deployment.extensions/my-first-demo-dep rolled back
查看回滚过程状态:
$ kubectl rollout status deployment my-first-demo-dep
Waiting for deployment "my-first-demo-dep" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "my-first-demo-dep" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "my-first-demo-dep" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "my-first-demo-dep" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "my-first-demo-dep" rollout to finish: 1 old replicas are pending termination...
deployment "my-first-demo-dep" successfully rolled out
上面的日志显示了完整的回滚过程,看到最后一行successfully rolled out
输出的时候,说明回滚完成了。这时候响应的应该就是 Hello world!
,我们来验证一下回滚效果:
$ curl 'http://localhost:30000/'
<h1>Hello world! <h1>
OK,成功了,回滚到了没有错误的原始版本。
总结
本篇演示了Kubenetes集群发布,集群副本集查看,集群自愈能力演示,集群扩容,滚动升级,以及回滚。然而这一切来得是如此的简单,我们只需要修改一下配置,执行一个命令,剩下的工作Kubenetes都帮我们做了。当我在回顾我几年前所在的某家公司的时候,程序升级上线都是由运维手工完成,虽然也完成了任务,但是那种原始的操作方式,在时下如果还有人用的话,那简直是不敢想象。
Docker & k8s 系列一:快速上手docker
Docker & k8s 系列二:本机k8s环境搭建
Docker & k8s 系列三:在k8s中部署单个服务实例