10、Kubernetes 实战 - 滚动发布的介绍与实现
一，前言
上一篇，介绍了灰度发布和流量切分的集中方式，以及如何实现 k8s 的灰度发布；
本篇，介绍滚动发布的实现；
二，滚动发布简介
滚动发布
-
滚动发布，则是我们一般所说的无宕机发布。其发布方式如同名称一样，一次取出一台/多台服务器（看策略配置）进行新版本更新。当取出的服务器新版确保无问题后，接着采用同等方式更新后面的服务器
-
k8s创建副本应用程序的最佳方法就是部署(Deployment)，部署自动创建副本集(ReplicaSet)，副本集可以精确地控制每次替换的Pod数量，从而可以很好的实现滚动更新
-
k8s每次使用一个新的副本控制器(replication controller)来替换已存在的副本控制器，从而始终使用一个新的Pod模板来替换旧的pod模板
-
创建一个新的 replication controller；
-
增加或减少 pod 副本数量，直到满足当前批次期望的数量；
-
删除掉旧的 replication controller；
发布流程和策略
-
优点
-
不需要停机更新，无感知平滑更新；
-
版本更新成本小,不需要新旧版本共存；
-
缺点
-
更新时间长：每次只更新一个/多个镜像，需要频繁连续等待服务启动缓冲
-
旧版本环境无法得到备份：始终只有一个环境存在
-
回滚版本异常：如果滚动发布到一半出了问题，回滚时需要使用同样的滚动策略回滚旧版本
三、环境初始化
由于k8s-node 资源问题，先清理一下 k8s-node 上的资源占用；
查看资源情况
[root@k8s-master deployment]# kubectl get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
nginx 0/3 3 0 2d1h
pay-v1 0/3 3 0 6h52m
user-v1 0/3 3 0 8m11s
user-v2 0/3 3 0 5h11m
[root@k8s-master deployment]# kubectl get service
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 2d16h
nginx NodePort 10.107.223.32 <none> 80:32117/TCP 2d1h
service-pay-v1 NodePort 10.106.98.218 <none> 80:30872/TCP 6h32m
service-user-v1 NodePort 10.104.13.40 <none> 80:31071/TCP 28h
service-user-v2 NodePort 10.100.196.142 <none> 80:30289/TCP 5h12m
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
mysql-g2zst 0/1 CrashLoopBackOff 53 2d
nginx-6799fc88d8-2wvl2 1/1 Running 1 2d1h
nginx-6799fc88d8-lkct4 1/1 Running 1 2d1h
nginx-6799fc88d8-pktqq 1/1 Running 1 2d1h
pay-v1-655587b6f5-lpft2 1/1 Running 1 6h41m
pay-v1-655587b6f5-pcnrp 1/1 Running 2 6h42m
pay-v1-655587b6f5-spj85 1/1 Running 1 6h41m
user-v1-5895c69847-8tkm9 0/1 Pending 0 9m18s
user-v1-5895c69847-swp52 0/1 Pending 0 9m18s
user-v1-5895c69847-xr4r8 0/1 Pending 0 9m18s
user-v2-fc9d84585-2zztd 1/1 Running 2 5h13m
user-v2-fc9d84585-ss2ss 1/1 Running 1 5h13m
user-v2-fc9d84585-xrvnf 1/1 Running 1 5h13m
删除无用 pod
注意：想要删除 pod，需要删除掉 deployment，否则副本依然还在；
以nginx 为例：目前的有 3 个正在运行的 nginx 副本
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
mysql-4nq2q 0/1 Pending 0 4m19s
nginx-6799fc88d8-2wvl2 1/1 Running 1 2d1h
nginx-6799fc88d8-lkct4 1/1 Running 1 2d2h
nginx-6799fc88d8-pktqq 1/1 Running 1 2d1h
pay-v1-655587b6f5-lpft2 1/1 Running 1 6h53m
pay-v1-655587b6f5-pcnrp 1/1 Running 2 6h53m
pay-v1-655587b6f5-spj85 1/1 Running 1 6h53m
user-v1-5895c69847-8tkm9 0/1 Pending 0 20m
user-v1-5895c69847-swp52 0/1 Pending 0 20m
user-v1-5895c69847-xr4r8 0/1 Pending 0 20m
user-v2-fc9d84585-2zztd 1/1 Running 2 5h24m
user-v2-fc9d84585-ss2ss 1/1 Running 1 5h24m
user-v2-fc9d84585-xrvnf 1/1 Running 1 5h24m
[root@k8s-master deployment]# kubectl get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
nginx 0/3 3 0 2d2h
pay-v1 0/3 3 0 7h8m
user-v1 0/3 3 0 23m
user-v2 0/3 3 0 5h27m
删除nginx 的 deploy 后，pod 状态变为 Terminating
[root@k8s-master deployment]# kubectl delete deployment nginx
deployment.apps "nginx" deleted
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
mysql-4nq2q 0/1 Pending 0 9m50s
nginx-6799fc88d8-2wvl2 1/1 Terminating 1 2d1h
nginx-6799fc88d8-lkct4 1/1 Terminating 1 2d2h
nginx-6799fc88d8-pktqq 1/1 Terminating 1 2d1h
pay-v1-655587b6f5-lpft2 1/1 Terminating 1 6h58m
pay-v1-655587b6f5-pcnrp 1/1 Terminating 2 6h59m
pay-v1-655587b6f5-spj85 1/1 Terminating 1 6h58m
user-v1-5895c69847-8tkm9 0/1 Pending 0 26m
user-v1-5895c69847-swp52 0/1 Pending 0 26m
user-v1-5895c69847-xr4r8 0/1 Pending 0 26m
user-v2-fc9d84585-2zztd 1/1 Terminating 2 5h30m
user-v2-fc9d84585-ss2ss 1/1 Terminating 1 5h30m
user-v2-fc9d84585-xrvnf 1/1 Terminating 1 5h30m
继续再删除 pod
[root@k8s-master deployment]# kubectl delete pod nginx-6799fc88d8-2wvl2 --force
warning: Immediate deletion does not wait for confirmation that the running resource has been terminated. The resource may continue to run on the cluster indefinitely.
pod "nginx-6799fc88d8-2wvl2" force deleted
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
mysql-4nq2q 0/1 Pending 0 14m
nginx-6799fc88d8-lkct4 1/1 Terminating 1 2d2h
nginx-6799fc88d8-pktqq 1/1 Terminating 1 2d1h
pay-v1-655587b6f5-lpft2 1/1 Terminating 1 7h3m
pay-v1-655587b6f5-pcnrp 1/1 Terminating 2 7h3m
pay-v1-655587b6f5-spj85 1/1 Terminating 1 7h3m
user-v1-5895c69847-8tkm9 0/1 Pending 0 30m
user-v1-5895c69847-swp52 0/1 Pending 0 30m
user-v1-5895c69847-xr4r8 0/1 Pending 0 30m
user-v2-fc9d84585-2zztd 1/1 Terminating 2 5h34m
user-v2-fc9d84585-ss2ss 1/1 Terminating 1 5h34m
user-v2-fc9d84585-xrvnf 1/1 Terminating 1 5h34m
重复操作，删除全部无用 pod
这个mysql 删除不掉
[root@k8s-master deployment]# kubectl get deploy
NAME READY UP-TO-DATE AVAILABLE AGE
user-v1 0/3 3 0 33m
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
mysql-xqsgz 0/1 Pending 0 13s
user-v1-5895c69847-8tkm9 0/1 Pending 0 35m
user-v1-5895c69847-swp52 0/1 Pending 0 35m
user-v1-5895c69847-xr4r8 0/1 Pending 0 35m
删除 Service
kubectl delete service service-pay-v1 service-user-v2 nginx
[root@k8s-master deployment]# kubectl get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 2d16h
service-user-v1 NodePort 10.104.13.40 <none> 80:31071/TCP 28h
重启阿里云服务
[root@k8s-master deployment]# kubectl get nodes
NAME STATUS ROLES AGE VERSION
k8s-master Ready control-plane,master 2d16h v1.20.4
k8s-node Ready <none> 2d16h v1.20.4
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
user-v1-5895c69847-8zs9n 1/1 Running 0 20m
user-v1-5895c69847-tdkhj 1/1 Running 0 20m
user-v1-5895c69847-xcdwl 1/1 Running 0 20m
测试连接
[root@k8s-master deployment]# curl http://172.17.178.106:31071
user-v1
[root@k8s-master deployment]# curl http://172.17.178.105:31071
user-v1
四，实现滚动更新
重新做一下，用 user-v1 测试滚动更新
目前基于 deployment-user-v1.yaml 创建了 deployment 部署实例；
基于user-service-v1.yaml 创建了 service，并创建了 3 个 pod 副本
修改 deployment 配置，实现滚动更新
将v1 升级成 v3，原始 deployment-user-v1.yaml：
[root@k8s-master deployment]# cat deployment-user-v1.yaml
apiVersion: apps/v1 API版本号
kind: Deployment 资源类型部署
metadata:
name: user-v1 资源名称
spec:
selector:
matchLabels:
app: user-v1 告诉deployment根据规则匹配相应的Pod进行控制和管理，matchLabels字段匹配Pod的label值
replicas: 3 声明Pod副本的数量
template:
metadata:
labels:
app: user-v1Pod名称
spec: 描述Pod内的容器信息
containers:
- name: nginx 容器的名称
image: nginx:user-v1镜像
ports:
- containerPort: 80容器内映射的端口
修改deployment-user-v1.yaml
apiVersion: apps/v1 API 配置版本
kind: Deployment 资源类型
metadata:
name: user-v1 资源名称
spec:
+ minReadySeconds: 1
+ strategy:
+ type: RollingUpdate
+ rollingUpdate:
+ maxSurge: 1
+ maxUnavailable: 0
+ selector:
+ matchLabels:
+ app: user-v1告诉deployment根据规则匹配相应的Pod进行控制和管理，matchLabels字段匹配Pod的label值
replicas: 10声明一个 Pod,副本的数量
template:
metadata:
labels:
app: user-v1Pod的名称
spec: 组内创建的 Pod 信息
containers:
- name: nginx容器的名称
+ image: nginx:user-v3使用哪个镜像
ports:
- containerPort: 80容器内映射的端口
相关参数介绍
| 参数 | 含义 | 备注 |
|---|---|---|
| minReadySeconds | 容器接受流量延缓时间：单位为秒，默认为0。如果没有设置的话，k8s会认为容器启动成功后就可以用了。设置该值可以延缓容器流量切分 | 容器启动好了，并不代表服务启动好了，比如还要去连接数据库，初始化缓存等服务初始化工作 |
| strategy.type = RollingUpdate | ReplicaSet 发布类型，声明为滚动发布，默认也为滚动发布 | |
| strategy.rollingUpdate.maxSurge | 最多Pod数量：为数字类型/百分比。如果 maxSurge 设置为1，replicas 设置为10，则在发布过程中pod数量最多为10 + 1个（多出来的为旧版本pod，平滑期不可用状态）。maxUnavailable 为 0 时，该值也不能设置为0 | |
| strategy.rollingUpdate.maxUnavailable | 升级中最多不可用pod的数量：为数字类型/百分比。当 maxSurge 为 0 时，该值也不能设置为0 |
扩容数量，是不包括激增量的
pod最大数量 = 容量 + 激增数
应用配置
[root@k8s-master deployment]# kubectl apply -f deployment-user-v1.yaml
deployment.apps/user-v1 configured
分析更新过程
// 更新前
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
user-v1-5895c69847-9mdm8 1/1 Running 0 2m49s
user-v1-5895c69847-wjzfh 1/1 Running 0 2m51s
user-v1-5895c69847-x8n7z 1/1 Running 0 2m54s
// 开始更新
[root@k8s-master deployment]# kubectl apply -f deployment-user-v1.yaml
deployment.apps/user-v1 configured
// 更新中
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
user-v1-5895c69847-9mdm8 1/1 Terminating 0 3m27s
user-v1-5895c69847-wjzfh 1/1 Running 0 3m29s
user-v1-5895c69847-x8n7z 1/1 Running 0 3m32s
user-v1-8cc9f4fb5-52hmd 0/1 ContainerCreating 0 0s
user-v1-8cc9f4fb5-zqj2l 1/1 Running 0 3s
// 更新完成
[root@k8s-master deployment]# kubectl get pods
NAME READY STATUS RESTARTS AGE
user-v1-8cc9f4fb5-52hmd 1/1 Running 0 48s
user-v1-8cc9f4fb5-6l7mz 1/1 Running 0 46s
user-v1-8cc9f4fb5-zqj2l 1/1 Running 0 51s
验证使用镜像是否变更：
[root@k8s-master deployment]# kubectl describe pod user-v1-8cc9f4fb5-zqj2l
Name: user-v1-8cc9f4fb5-zqj2l
Namespace: default
Priority: 0
Node: k8s-node/172.17.178.106
Start Time: Fri, 24 Dec 2021 18:30:57 +0800
Labels: app=user-v1
pod-template-hash=8cc9f4fb5
Annotations: <none>
Status: Running
IP: 10.244.1.67
IPs:
IP: 10.244.1.67
Controlled By: ReplicaSet/user-v1-8cc9f4fb5
Containers:
nginx:
Container ID: docker://436d2058c9a6a85000b2c32a3a87c1b7c892427894db1c7d851adb953ba01e88
Image: nginx:user-v3
Image ID: nginx@sha256:c3826e7a5cd2abc9f0e92fe02fe4897ab4346539438ce4f6c811371677f7494b
Port: 80/TCP
Host Port: 0/TCP
State: Running
Started: Fri, 24 Dec 2021 18:30:58 +0800
Ready: True
Restart Count: 0
Environment: <none>
Mounts:
/var/run/secrets/kubernetes.io/serviceaccount from default-token-q4qxd (ro)
Conditions:
Type Status
Initialized True
Ready True
ContainersReady True
PodScheduled True
Volumes:
default-token-q4qxd:
Type: Secret (a volume populated by a Secret)
SecretName: default-token-q4qxd
Optional: false
QoS Class: BestEffort
Node-Selectors: <none>
Tolerations: node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Scheduled 7m8s default-scheduler Successfully assigned default/user-v1-8cc9f4fb5-zqj2l to k8s-node
Normal Pulled 7m8s kubelet Container image "nginx:user-v3" already present on machine
Normal Created 7m8s kubelet Created container nginx
Normal Started 7m8s kubelet Started container nginx
是nginx:user-v3，说明应用已更新成功；
访问 service-user-v1
curl http://172.17.178.105:31071
curl http://172.17.178.106:31071
[root@k8s-master deployment]# curl http://172.17.178.105:31071
user-v3
[root@k8s-master deployment]# curl http://172.17.178.106:31071
user-v3
五，结尾
下一篇，服务的可用性探针；