Joe Blog
什麼是Argo Rollouts
Argo Rollouts 是一个Kubernetes 控制器和一组CRD，为Kubernetes提供高级部署功能，例如蓝绿部署、金丝雀部署、金丝雀分析、更新测试(experimentation)和渐进式交付(progressive delivery)功能。
Argo Rollouts可与入口控制器和服务网格集成，利用它们的流量整形能力在更新期间逐渐将流量转移到新版本。此外，Argo Rollouts 可以查询的各種providers的Metrics指标（可定制的metric查询和kpi分析），并在更新期间進行自动升级或回滚。
原理
Argo Rollouts原理和Deployment的差不多，其CRD定義及源碼引用了Deployment一部分數據結構，只是加强rollout的策略和流量控制。当spec.template发送变化时，Argo-Rollout就会根据spec.strategy进行rollout，通常会产生一个新的ReplicaSet，逐步scale down之前的ReplicaSet的pod数量。
安裝
# instal argo rollouts controller and CRD
kubectl create namespace argo-rollouts
kubectl apply -n argo-rollouts -f https://github.com/argoproj/argo-rollouts/releases/latest/download/install.yaml
# install argo rollouts kubectl plugin
curl -LO https://github.com/argoproj/argo-rollouts/releases/latest/download/kubectl-argo-rollouts-linux-amd64
chmod +x ./kubectl-argo-rollouts-linux-amd64
mv ./kubectl-argo-rollouts-linux-amd64 /usr/local/bin/kubectl-argo-rollouts
Argo Rollouts 部署策略
要使用 Argo Rollouts，我们声明一个资源，其apiVersion属性为 argoproj.io/v1alpha1，Kind为Rollout，如下所示：
apiVersion: argoproj.io/v1alpha1
kind: Rollout
Argo Rollouts 的配置有一个strategy属性供我们选择我们想要的部署策略，有两个blueGreen和canary。
在這裏我們先來了解blueGreen部署策略（藍/綠部署）。
實踐
blueGreen-rollouts.yaml
# This example demonstrates a Rollout using the blue-green update strategy, which contains a manual
# gate before promoting the new stack.
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
name: rollout-bluegreen
spec:
replicas: 2
revisionHistoryLimit: 2
selector:
matchLabels:
app: rollout-bluegreen
template:
metadata:
labels:
app: rollout-bluegreen
spec:
containers:
- name: rollouts-demo
image: argoproj/rollouts-demo:blue
imagePullPolicy: Always
ports:
- containerPort: 8080
strategy:
blueGreen:
# activeService specifies the service to update with the new template hash at time of promotion.
# This field is mandatory for the blueGreen update strategy.
activeService: rollout-bluegreen-active
# previewService specifies the service to update with the new template hash before promotion.
# This allows the preview stack to be reachable without serving production traffic.
# This field is optional.
previewService: rollout-bluegreen-preview
# autoPromotionEnabled disables automated promotion of the new stack by pausing the rollout
# immediately before the promotion. If omitted, the default behavior is to promote the new
# stack as soon as the ReplicaSet are completely ready/available.
# Rollouts can be resumed using: `kubectl argo rollouts promote ROLLOUT`
autoPromotionEnabled: false
---
kind: Service
apiVersion: v1
metadata:
name: rollout-bluegreen-active
spec:
type: NodePort
selector:
app: rollout-bluegreen
ports:
- protocol: TCP
port: 80
targetPort: 8080
---
kind: Service
apiVersion: v1
metadata:
name: rollout-bluegreen-preview
spec:
type: NodePort
selector:
app: rollout-bluegreen
ports:
- protocol: TCP
port: 80
targetPort: 8080
Rollout 的數據結構与原生的 Deployment Object 相同，只是strategy属性不同。
在上面的文件中，我们声明了 3 个有關於blueGreen 的strategy属性：
autoPromotionEnabled: false— 表示 rollout 是否应自动将新的 ReplicaSet 提升为Active或进入Pause状态。如果未指定，则默认值为 true。activeService: brollout-bluegreen-active— 对部署修改为Active服务的服务的引用。previewService: rollout-bluegreen-preview— 推出修改为预览服务的服务的名称。
以上yaml中创建2個service，它們除了name属性之外，这两个 Services的selector屬性是一致的。
...
selector:
app: rollout-bluegreen
...
创建一个 Rollout 对象和二個 Service 服務
root@node1:~/argo-rollouts# kubectl apply -f bluegreen-rollout.yaml -n argo-rollouts
rollout.argoproj.io/rollout-bluegreen created
service/rollout-bluegreen-active created
service/rollout-bluegreen-preview created
root@node1:~/argo-rollouts# kubectl get svc -n argo-rollouts
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
argo-rollouts-dashboard ClusterIP 10.233.46.185 <none> 3100/TCP 2d1h
argo-rollouts-metrics ClusterIP 10.233.42.13 <none> 8090/TCP 2d5h
rollout-bluegreen-active NodePort 10.233.15.20 <none> 80:32677/TCP 3m41s
rollout-bluegreen-preview NodePort 10.233.15.157 <none> 80:31906/TCP 3m40s
root@node1:~/argo-rollouts#
root@node1:~/argo-rollouts# kubectl get rollout -n argo-rollouts
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
rollout-bluegreen 2 2 2 2 3m52s
root@node1:~/ephemeral-containers# kubectl exec -it busybox -n default /bin/sh
kubectl exec [POD] [COMMAND] is DEPRECATED and will be removed in a future version. Use kubectl exec [POD] -- [COMMAND] instead.
/ # nslookup rollout-bluegreen-active.argo-rollouts.svc.cluster.local.
Server: 169.254.25.10
Address 1: 169.254.25.10
Name: rollout-bluegreen-active.argo-rollouts.svc.cluster.local.
Address 1: 10.233.15.20 rollout-bluegreen-active.argo-rollouts.svc.cluster.local
/ # nslookup rollout-bluegreen-preview.argo-rollouts.svc.cluster.local.
Server: 169.254.25.10
Address 1: 169.254.25.10
Name: rollout-bluegreen-preview.argo-rollouts.svc.cluster.local.
Address 1: 10.233.15.157 rollout-bluegreen-preview.argo-rollouts.svc.cluster.local
root@node1:~/ephemeral-containers# kubectl get rs -n argo-rollouts
NAME DESIRED CURRENT READY AGE
argo-rollouts-76fcfc8d7f 1 1 1 2d6h
argo-rollouts-dashboard-75f845ff57 1 1 1 2d1h
rollout-bluegreen-5ffd47b8d4 2 2 2 27m
root@node1:~/ephemeral-containers# kubectl get pod -n argo-rollouts -owide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
argo-rollouts-76fcfc8d7f-tnb9w 1/1 Running 0 2d6h 10.233.67.164 node5 <none> <none>
argo-rollouts-dashboard-75f845ff57-kw64d 1/1 Running 0 2d1h 10.233.68.186 node4 <none> <none>
rollout-bluegreen-5ffd47b8d4-cp4ws 1/1 Running 0 25m 10.233.67.213 node5 <none> <none>
rollout-bluegreen-5ffd47b8d4-nb7mh 1/1 Running 0 25m 10.233.68.79 node4 <none> <none>
root@node1:~/ephemeral-containers# kubectl get ep -n argo-rollouts -owide
NAME ENDPOINTS AGE
argo-rollouts-dashboard 10.233.68.186:3100 2d1h
argo-rollouts-metrics 10.233.67.164:8090 2d6h
rollout-bluegreen-active 10.233.67.213:8080,10.233.68.79:8080 28m
rollout-bluegreen-preview 10.233.67.213:8080,10.233.68.79:8080 28m
此时，rollout-bluegreen-active和rollout-bluegreen-preview都指向
同一個rollout-bluegreen-5ffd47b8d4的ReplicaSet。
打開瀏覽器
訪問rollout-bluegreen-active和rollout-bluegreen-preview的UI 發現是一樣的服務
现在，我们更改imageRollout Containers Spec image的tag的属性：
...
spec:
containers:
- name: rollouts-demo
image: argoproj/rollouts-demo:green
...
更新yaml文件apply
root@node1:~/argo-rollouts# kubectl -n argo-rollouts apply -f bluegreen-rollout.yaml
rollout.argoproj.io/rollout-bluegreen configured
service/rollout-bluegreen-active unchanged
service/rollout-bluegreen-preview unchanged
#此时，Argo Rollouts 将为新yaml文件创建一个新的ReplicaSet: rollout-bluegreen-75695867f
root@node1:~/argo-rollouts# kubectl -n argo-rollouts get rs
NAME DESIRED CURRENT READY AGE
argo-rollouts-76fcfc8d7f 1 1 1 2d7h
argo-rollouts-dashboard-75f845ff57 1 1 1 2d2h
rollout-bluegreen-5ffd47b8d4 2 2 2 85m
rollout-bluegreen-75695867f 2 2 2 74s
#然后将rollout-bluegreen-preview服务修改为指向新的 ReplicaSet,對應的endppoint地址有變化
root@node1:~/argo-rollouts# kubectl -n argo-rollouts get ep
NAME ENDPOINTS AGE
argo-rollouts-dashboard 10.233.68.186:3100 2d2h
argo-rollouts-metrics 10.233.67.164:8090 2d7h
rollout-bluegreen-active 10.233.67.213:8080,10.233.68.79:8080 86m
rollout-bluegreen-preview 10.233.66.4:8080,10.233.69.98:8080 86m
访问rollout-bluegreen-preview的不同的UI
而rollout-bluegreen-active服务UI没有变化
使用Argo-Rollout提供的plugin查看其状态：
root@node1:~/argo-rollouts# kubectl argo rollouts get rollout rollout-bluegreen -n argo-rollouts
Name: rollout-bluegreen
Namespace: argo-rollouts
Status: ॥ Paused
Message: BlueGreenPause
Strategy: BlueGreen
Images: argoproj/rollouts-demo:blue (stable, active)
argoproj/rollouts-demo:green (preview)
Replicas:
Desired: 2
Current: 4
Updated: 2
Ready: 2
Available: 2
NAME KIND STATUS AGE INFO
⟳ rollout-bluegreen Rollout ॥ Paused 98m
├──# revision:2
│ └──⧉ rollout-bluegreen-75695867f ReplicaSet ✔ Healthy 14m preview
│ ├──□ rollout-bluegreen-75695867f-mhjnv Pod ✔ Running 14m ready:1/1
│ └──□ rollout-bluegreen-75695867f-mng65 Pod ✔ Running 14m ready:1/1
└──# revision:1
└──⧉ rollout-bluegreen-5ffd47b8d4 ReplicaSet ✔ Healthy 98m stable,active
├──□ rollout-bluegreen-5ffd47b8d4-cp4ws Pod ✔ Running 98m ready:1/1
└──□ rollout-bluegreen-5ffd47b8d4-nb7mh Pod ✔ Running 98m ready:1/1
接下来，我们通过promote來把rollout-bluegreen服务指向它来 Rollout到ReplicaSet的新版本，我们运行以下命令
root@node1:~/argo-rollouts# kubectl argo rollouts promote rollout-bluegreen -n argo-rollouts
rollout 'rollout-bluegreen' promoted
使用Argo-Rollout提供的plugin查看其状态：
root@node1:~/argo-rollouts# kubectl argo rollouts get rollout rollout-bluegreen -n argo-rollouts
Name: rollout-bluegreen
Namespace: argo-rollouts
Status: ✔ Healthy
Strategy: BlueGreen
Images: argoproj/rollouts-demo:blue
argoproj/rollouts-demo:green (stable, active)
Replicas:
Desired: 2
Current: 4
Updated: 2
Ready: 2
Available: 2
NAME KIND STATUS AGE INFO
⟳ rollout-bluegreen Rollout ✔ Healthy 105m
├──# revision:2
│ └──⧉ rollout-bluegreen-75695867f ReplicaSet ✔ Healthy 20m stable,active
│ ├──□ rollout-bluegreen-75695867f-mhjnv Pod ✔ Running 20m ready:1/1
│ └──□ rollout-bluegreen-75695867f-mng65 Pod ✔ Running 20m ready:1/1
└──# revision:1
└──⧉ rollout-bluegreen-5ffd47b8d4 ReplicaSet ✔ Healthy 104m delay:8s
├──□ rollout-bluegreen-5ffd47b8d4-cp4ws Pod ✔ Running 104m ready:1/1
└──□ rollout-bluegreen-5ffd47b8d4-nb7mh Pod ✔ Running 104m ready:1/1
訪問rollout-bluegreen-active和rollout-bluegreen-preview的UI 發現是一樣的服務
现在，Argo Rollouts 更新rollout-bluegreen服务以指向新的 ReplicaSet，等待（默认 30 秒）后，旧的 ReplicaSet 被缩小
root@node1:~/argo-rollouts# kubectl argo rollouts get rollout rollout-bluegreen -n argo-rollouts
Name: rollout-bluegreen
Namespace: argo-rollouts
Status: ✔ Healthy
Strategy: BlueGreen
Images: argoproj/rollouts-demo:green (stable, active)
Replicas:
Desired: 2
Current: 2
Updated: 2
Ready: 2
Available: 2
NAME KIND STATUS AGE INFO
⟳ rollout-bluegreen Rollout ✔ Healthy 112m
├──# revision:2
│ └──⧉ rollout-bluegreen-75695867f ReplicaSet ✔ Healthy 27m stable,active
│ ├──□ rollout-bluegreen-75695867f-mhjnv Pod ✔ Running 27m ready:1/1
│ └──□ rollout-bluegreen-75695867f-mng65 Pod ✔ Running 27m ready:1/1
└──# revision:1
└──⧉ rollout-bluegreen-5ffd47b8d4 ReplicaSet • ScaledDown 111m
用户界面Dashboard
可以在Dashboard裏做之前命令行進行的Promote
结论
用户希望在新版本开始为生产流量提供服务之前对其运行最后一分钟的功能测试。借助 BlueGreen 策略，Argo Rollouts 允许用户指定预览服务和活动服务。Argo Rollouts 将配置预览服务以将流量发送到新版本，而活动服务继续接收生产流量。一旦用户满意，他们可以将预览服务推广为新的活动服务。 本文简单体验了一下Argo Rollouts的蓝/绿部署功能，学会了如何使用 Argo Rollouts 自动化蓝/绿部署，正如你所看到的，它也很简单。