服务管理组件-Serving

Posted 2022-02-24 gg888666

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了服务管理组件-Serving相关的知识，希望对你有一定的参考价值。

Knative工作模式

service

　　route RevisionX

　　configuration

IngressGateway(Envoy) 收发流量

Revision**　　

　　k8s-Service Avtivator(0) SKS Deploument Pod(Envoy Queue container) 拉取指标到 Autoscaler HPA

Route**

Ingress(Istio) VirtualService

Service-to-service Pod(envoy Queue container)

Knative Serving会为每个Pod注入一个称为Queue Proxy的容器

◼ 为业务代码提供代理功能

　　◆Ingress GW接收到请求后，将其发往目标Pod实例上由Queue Proxy监听的8012端口

　　◆而后，Queue Proxy再将请求转发给业务代码容器监听的端口

◼ 报告实例上的多个有关客户端请求的关键指标为给Autoscaler

◼ Pod健康状态检测（Kubernetes探针）

◼ 为超出实例最大并发量的请求提供缓冲队列

为超出实例最大并发量的请求提供缓冲队列

Queue Proxy预留使用了如下几个端口

　　◆ 8012：HTTP协议的代理端口

　　◆ 8013：HTTP/2端口，用于代理gRPC

　　◆ 8022：管理端口，如健康状态检测等

　　◆ 9090：暴露给Autoscaler进行指标采集的端口

　　◆ 9091：暴露给Prometheus进行监控指标采集的端口

运行Knative应用

在Serving上，可通过创建Knative Service对象来运行应用；该Service资源会自动创建

◼ 一个Configuration对象，它会创建一个Revision，并由该Revision自动创建如下两个对象

　　◆一个Deployment对象

　　◆一个PodAutoscaler对象

◼ 一个Route对象，它会创建

　　◆一个Kubernetes Service对象

　　◆一组Istio VirtualService对象

　　　　⚫ kservice-name-ingress

　　　　⚫ kservice-name-mesh

Knative Service资源（简称kservice或ksvc）的资源规范主要有两个字段

◼ template：用于创建或更新Configuration，任何更新，都将创建新的Revision对象

◼ traffic：用于创建或更新Route对象

root@master01:/opt/knative-in-practise# kubectl get svc -nistio-system
NAME                    TYPE           CLUSTER-IP      EXTERNAL-IP      PORT(S)                                      AGE
istio-egressgateway     ClusterIP      10.100.141.15   <none>           80/TCP,443/TCP                               12d
istio-ingressgateway    LoadBalancer   10.100.43.136   192.168.80.251   15021:36049/TCP,80:29931/TCP,443:25555/TCP   12d  #南北向流量
istiod                  ClusterIP      10.100.35.2     <none>           15010/TCP,15012/TCP,443/TCP,15014/TCP        12d
knative-local-gateway   ClusterIP      10.100.5.126    <none>           80/TCP                                       45h  #东西向流量

root@master01:/opt/knative-in-practise/serving/basic# cat hello-world.yaml 
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: hello
spec:
  template:
    metadata:
      # This is the name of our new "Revision," it must follow the convention service-name-revision-name
      name: hello-world
    spec:
      containers:
        #- image: gcr.io/knative-samples/helloworld-go
        - image: ikubernetes/helloworld-go:latest
          ports:
            - containerPort: 8080
          env:
            - name: TARGET
              value: "World"

链接：https://pan.baidu.com/s/1XrIcxCkDwy0hggoH1wqrgg

root@master01:~# kubectl get configurations,revisions,deployments,podautoscalers
NAME                                      LATESTCREATED   LATESTREADY   READY   REASON
configuration.serving.knative.dev/hello   hello-world     hello-world   True    

NAME                                       CONFIG NAME   K8S SERVICE NAME   GENERATION   READY   REASON   ACTUAL REPLICAS   DESIRED REPLICAS
revision.serving.knative.dev/hello-world   hello                            1            True             1                 1

NAME                                     READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/hello-world-deployment   1/1     1            1           10m


NAME                                                         DESIREDSCALE   ACTUALSCALE   READY   REASON
podautoscaler.autoscaling.internal.knative.dev/hello-world   1              1             True

root@master01:~# kubectl get routes,services,virtualservice
NAME                              URL                                READY   REASON
route.serving.knative.dev/hello   http://hello.default.example.com   True    

NAME                          TYPE           CLUSTER-IP      EXTERNAL-IP                                            PORT(S)                                      AGE
service/demoappv10            ClusterIP      10.100.20.78    <none>                                                 8080/TCP                                     13d
service/hello                 ExternalName   <none>          knative-local-gateway.istio-system.svc.cluster.local   80/TCP                                       7m34s
service/hello-world           ClusterIP      10.100.87.121   <none>                                                 80/TCP                                       8m43s
service/hello-world-private   ClusterIP      10.100.186.63   <none>                                                 80/TCP,9090/TCP,9091/TCP,8022/TCP,8012/TCP   8m43s
service/kubernetes            ClusterIP      10.100.0.1      <none>                                                 443/TCP                                      13d
service/proxy                 ClusterIP      10.100.225.55   <none>                                                 80/TCP                                       13d

NAME                                               GATEWAYS                                                                              HOSTS                                                                                                 AGE
virtualservice.networking.istio.io/hello-ingress   ["knative-serving/knative-ingress-gateway","knative-serving/knative-local-gateway"]   ["hello.default","hello.default.example.com","hello.default.svc","hello.default.svc.cluster.local"]   7m34s
virtualservice.networking.istio.io/hello-mesh      ["mesh"]                                                                              ["hello.default","hello.default.svc","hello.default.svc.cluster.local"]

kubectl edit  deployment.apps/hello-world-deployment 
image: gcr.io/knative-releases/knative.dev/serving/cmd/queue:latest

root@master01:~# curl -H "Host: hello.default.example.com" 192.168.80.251
Hello World!

root@master01:~# kn service list
NAME    URL                                LATEST        AGE   CONDITIONS   READY   REASON
hello   http://hello.default.example.com   hello-world   80s   3 OK / 3     True

root@master01:~# kn service describe hello
Name:       hello
Namespace:  default
Age:        2m
URL:        http://hello.default.example.com

Revisions:  
  100%  @latest (hello-world) [1] (2m)
        Image:     ikubernetes/helloworld-go:latest (at 5ea96b)
        Replicas:  1/1

Conditions:  
  OK TYPE                   AGE REASON
  ++ Ready                   1m 
  ++ ConfigurationsReady     1m 
  ++ RoutesReady             1m

root@master01:~# kn revision list
NAME          SERVICE   TRAFFIC   TAGS   GENERATION   AGE     CONDITIONS   READY   REASON
hello-world   hello     100%             1            3m30s   4 OK / 4     True    
root@master01:~# kn revision describe hello-world
Name:       hello-world
Namespace:  default
Age:        3m
Image:      ikubernetes/helloworld-go:latest (at 5ea96b)
Replicas:   1/1
Port:       8080
Env:        TARGET=World
Service:    hello

Conditions:  
  OK TYPE                  AGE REASON
  ++ Ready                  3m 
  ++ ContainerHealthy       3m 
  ++ ResourcesAvailable     3m 
  ++ Active                 3m

root@master01:~# kn route list
NAME    URL                                READY
hello   http://hello.default.example.com   True
root@master01:~# kn route describe hello
Name:       hello
Namespace:  default
Age:        5m
URL:        http://hello.default.example.com
Service:    hello

Traffic Targets:  
  100%  @latest (hello-world)

Conditions:  
  OK TYPE                      AGE REASON
  ++ Ready                      4m 
  ++ AllTrafficAssigned         4m 
  ++ CertificateProvisioned     4m TLSNotEnabled
  ++ IngressReady               4m

root@master01:~# kn --help

root@master01:~# kn route list
NAME    URL                                READY
hello   http://hello.default.example.com   True
root@master01:~# kn domain create --help
root@master01:~# kn domain create hello.chuan.com --ref ksvc:hello  #创建域名映射
Domain mapping \'hello.chuan.com\' created in namespace \'default\'.

root@master01:~# kubectl get domainmapping
NAME              URL                      READY   REASON
hello.chuan.com   http://hello.magedu.com   False   DomainAlreadyClaimed
root@master01:~# kn domain describe hello.magedu.com
Name:       hello.chuan.com
Namespace:  default
Age:        4m

URL:  http://hello.magedu.com

Reference:     
  APIVersion:  serving.knative.dev/v1
  Kind:        Service
  Name:        hello

Conditions:  
  OK TYPE                      AGE REASON
  !! Ready                      4m DomainAlreadyClaimed
  ?? CertificateProvisioned     4m 
  !! DomainClaimed              4m DomainAlreadyClaimed
  ?? IngressReady               4m IngressNotConfigured
  ?? ReferenceResolved          4m

root@master01:/opt/knative-in-practise/serving/domainmapping# cat  cdc-hello.magedu.com.yaml   #域名映射需调整比列
apiVersion: networking.internal.knative.dev/v1alpha1
kind: ClusterDomainClaim
metadata:
  name: hello.magedu.com
spec:
  namespace: default

root@master01:/opt/knative-in-practise/serving/domainmapping# kubectl apply -f cdc-hello.magedu.com.yaml

root@master01:/opt/knative-in-practise/serving/domainmapping# kubectl get cdc
NAME               AGE
hello.magedu.com   16s

root@master01:/opt/knative-in-practise/serving/domainmapping# kubectl get vs
NAME                       GATEWAYS                                                                              HOSTS                                                                                                 AGE
hello-ingress              ["knative-serving/knative-ingress-gateway","knative-serving/knative-local-gateway"]   ["hello.default","hello.default.example.com","hello.default.svc","hello.default.svc.cluster.local"]   43m
hello-mesh                 ["mesh"]                                                                              ["hello.default","hello.default.svc","hello.default.svc.cluster.local"]                               43m
hello.magedu.com-ingress   ["knative-serving/knative-ingress-gateway"]                                           ["hello.magedu.com"]

域名映射自动

root@master01:~# kubectl get cm  -nknative-serving
kubectl edit cm config-network  -nknative-serving  #
    autocreate-cluster-domain-claims: "false" 
  autocreate-cluster-domain-claims: "true"  #缩进两格，自动创建域名映射机制

root@master01:~# kn domain create hi.chuan.com --ref kroute:hello
Domain mapping \'hi.chuan.com\' created in namespace \'default\'.

root@master01:~# while true;do curl -H "Host: hi.chuan.com" 192.168.80.251;sleep 1;done
Hello World!

queue

root@master01:/opt/knative-in-practise/serving/basic# kubectl apply -f hello-world-002.yaml 
service.serving.knative.dev/hello configured
root@master01:/opt/knative-in-practise/serving/basic# cat hello-world-002.yaml 
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: hello
spec:
  template:
    metadata:
      # This is the name of our new "Revision," it must follow the convention service-name-revision-name
      name: hello-world-002
    spec:
      containerConcurrency: 10
      containers:
        #- image: gcr.io/knative-samples/helloworld-go
        - image: ikubernetes/helloworld-go:latest
          ports:
            - containerPort: 8080
          env:
            - name: TARGET
              value: "World-002"
root@client /# while true;do curl -H "Host: hello.magedu.com" 192.168.80.251;sleep 1;done
Hello World-002!
Hello World-002!

root@master01:/opt/knative-in-practise/serving/basic# kn service list
NAME    URL                                LATEST            AGE   CONDITIONS   READY   REASON
hello   http://hello.default.example.com   hello-world-002   63m   3 OK / 3     True    
root@master01:/opt/knative-in-practise/serving/basic# kn route list
NAME    URL                                READY
hello   http://hello.default.example.com   True
root@master01:/opt/knative-in-practise/serving/basic# kubectl get svc
hello                     ExternalName   <none>          knative-local-gateway.istio-system.svc.cluster.local   80/TCP                                       62m
hello-world               ClusterIP      10.100.38.120   <none>                                                 80/TCP                                       63m
hello-world-002           ClusterIP      10.100.9.130    <none>                                                 80/TCP                                       5m15s
hello-world-002-private   ClusterIP      10.100.23.65    <none>                                                 80/TCP,9090/TCP,9091/TCP,8022/TCP,8012/TCP   5m15s

调整比例

root@master01:/opt/knative-in-practise/serving/basic# cat hello-world.yaml 
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: hello
spec:
  template:
    metadata:
      # This is the name of our new "Revision," it must follow the convention service-name-revision-name
      name: hello-world
    spec:
      containers:
        #- image: gcr.io/knative-samples/helloworld-go
        - image: ikubernetes/helloworld-go:latest
          ports:
            - containerPort: 8080
          env:
            - name: TARGET
              value: "World"
root@master01:/opt/knative-in-practise/serving/basic# cat he.yaml 
apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: he
spec:
  template:
    metadata:
      # This is the name of our new "Revision," it must follow the convention service-name-revision-name
      name: he-world
    spec:
      containers:
        #- image: gcr.io/knative-samples/helloworld-go
        - image: ikubernetes/helloworld-go:latest
          ports:
            - containerPort: 8080
          env:
            - name: TARGET
              value: "chuan"

root@master01:/opt/knative-in-practise/serving/revision-and-route# cat 004-hello-CloudNative.yaml

apiVersion: serving.knative.dev/v1
kind: Service
metadata:
  name: fenge
spec:
  template:
    spec:
      containers:
        - image: ikubernetes/helloworld-go:latest
          ports:
            - containerPort: 8080
          env:
            - name: TARGET
              value: "Cloud-Native"
  traffic:
  - latestRevision: true
    percent: 0
    tag: staging
  - revisionName: he-world  #分别50%
    percent: 50
  - revisionName: hello-world
    percent: 50

root@master01:/opt/knative-in-practise/serving/basic# kn route list
NAME    URL                                READY
fenge   http://fenge.default.example.com   True
he      http://he.default.example.com      True
hello   http://hello.default.example.com   True
root@master01:/opt/knative-in-practise/serving/basic# kn revision list
NAME          SERVICE   TRAFFIC   TAGS      GENERATION   AGE     CONDITIONS   READY   REASON
fenge-00001   fenge               staging   1            3m9s    3 OK / 4     True    
he-world      he        100%                1            11m     4 OK / 4     True    
hello-world   hello     100%                1            5m10s   4 OK / 4     True

root@nt888 /# while true;do curl -H "Host: fenge.default.example.com" 192.168.80.251;sleep 1;done
Hello chuan!
Hello chuan!
Hello World!
Hello World!
Hello chuan!
Hello World!
Hello chuan!

kubectl get vs
staging-fenge.default.svc.cluster.local

root@nt888 /# curl http://staging-fenge.default.svc.cluster.local
Hello Cloud-Native!
root@nt888 /# curl http://staging-fenge.default.svc.cluster.local

增加标签

root@master01:~# kn service update fenge --tag fenge-00001=oldet
Updating Service \'fenge\' in namespace \'default\':

  0.033s The Route is still working to reflect the latest desired specification.
  0.075s Ingress has not yet been reconciled.
  0.138s Waiting for load balancer to be ready
  0.337s Ready to serve.

Service \'fenge\' with latest revision \'fenge-00001\' (unchanged) is available at URL:
http://fenge.default.example.com

root@nt888 /# while true;do curl http://oldet-fenge.default.svc.cluster.local;sleep 1;done
Hello Cloud-Native!
Hello Cloud-Native!
Hello Cloud-Native!

root@master01:~# kn revision list
NAME          SERVICE   TRAFFIC   TAGS            GENERATION   AGE    CONDITIONS   READY   REASON
fenge-00001   fenge               staging,oldet   1            8m4s   4 OK / 4     True    
he-world      he        100%                      1            75m    4 OK / 4     True    
hello-world   hello     100%                      1            68m    4 OK / 4     True

以上是关于服务管理组件-Serving的主要内容，如果未能解决你的问题，请参考以下文章

Knative 基本功能深入剖析：Knative Serving 之服务路由管理

Tensorflow Serving介绍及部署安装

【Knative系列】理解 Knative Serving扩缩容系统的设计

使用 TensorFlow Serving 服务 Keras 模型

为 TensorFlow Serving 保存模型

模型导出与部署TF Serving 与 Web开启服务