serverless开发

1.总得扯点啥

目前对kubernetes上无服务器架构(Faas/Serverless)都是热门话题，各大公司也都相继进行投入。本文主要是想通过实际例子，了解无服务器架构。

2.什么是无服务器架构(Faas/Serverless)

自己的认识: care code, don`t care server 。开发人员将通用的功能以函数的形式提交注册到kubernetes作为其中的CRD，使用这根据输入规范使用CRD。当然函数提供方和基础设施（kubernetes）提供方会按照相应的规则计量计费。

3.Serverless的开发框架

Serverless的开发框架由很多，比如：

• Kubeless ：Kubernetes原生的无服务器框架
• Fission ：Kubernetes的快速无服务器框架，专注于开发人员的生产力和高性能。支持ython，NodeJS，Go，C＃或PHP编写函数。
• Funktion ：专为Kubernetes设计的开源事件驱动的Lambda风格编程模型。Funktion与fabric8平台紧密相连。
• IronFunctions ：一个开源的无服务器平台或FaaS平台，可以在任何地方运行。IronFunction是用Golang编写的，实际上支持任何语言的功能。
  还有OpenWhisk，OpenFaaS，Nuclio，Virtual Kubelet，Fnproject等其他的开源框架。

4.以Kubeless为例，开发Serverless。

环境准备

# 下载kubeless
[root@node4 kubeless]# ll
total 9792
-rw-r--r-- 1 root root 10004863 Oct 14 14:47 kubeless_linux-amd64.zip
-rw-r--r-- 1 root root    17544 Oct 14 14:47 kubeless-v1.0.4.yaml
#解压将kubeless拷贝到/usr/local/bin
#如果不能俩连外网，预先加载以下镜像
#这里只下载go-init和go的镜像，因为这里的函数式基于golang写的，不同的开发语言可以下载对应的镜像
[root@node4 soft]# docker images
REPOSITORY                            TAG                 IMAGE ID            CREATED             SIZE
kubeless/function-controller          v1.0.4              b65aa9b834b7        3 months ago        85.3MB
kubeless/go-init                      latest              e07ff96bf0a5        3 months ago        954MB
kubeless/go                           latest              493ee8f547df        3 months ago        58.4MB
kubeless/http-trigger-controller      v1.0.1              d6f09f3299d9        3 months ago        83.4MB
kubeless/cronjob-trigger-controller   v1.0.1              aec1dd30bb57        3 months ago        77.1MB
kubeless/unzip                        latest              abb7a6ac3e05        3 months ago        75.4MB
bitnami/kubeless-ui                   latest              43d3109c8a74        9 months ago        231MB
lachlanevenson/k8s-kubectl            v1.13.1             d2525e8609a3        10 months ago       44.2MB
k8s.gcr.io/pause                      3.1                 da86e6ba6ca1        22 months ago       742kB

注册kubeless到kubernetes集群中

[root@node4 kubeless]# kubectl create ns kubeless
namespace/kubeless created
#下载地址：https://github.com/kubeless/kubeless/releases
[root@node4 kubeless]# kubectl apply -f kubeless-v1.0.4.yaml 
clusterrole.rbac.authorization.k8s.io/kubeless-controller-deployer created
clusterrolebinding.rbac.authorization.k8s.io/kubeless-controller-deployer created
customresourcedefinition.apiextensions.k8s.io/functions.kubeless.io created
customresourcedefinition.apiextensions.k8s.io/httptriggers.kubeless.io created
customresourcedefinition.apiextensions.k8s.io/cronjobtriggers.kubeless.io created
configmap/kubeless-config created
deployment.apps/kubeless-controller-manager created
serviceaccount/controller-acct created

#查看部署情况
[root@node4 kubeless]#  kubectl get all -n kubeless
NAME                                               READY   STATUS    RESTARTS   AGE
pod/kubeless-controller-manager-5bcb6757d9-wb6hr   3/3     Running   0          28s

NAME                                          READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/kubeless-controller-manager   1/1     1            1           29s

NAME                                                     DESIRED   CURRENT   READY   AGE
replicaset.apps/kubeless-controller-manager-5bcb6757d9   1         1         1       29s
#查看CRD
[root@node4 kubeless]# kubectl get crd |grep kubeless
cronjobtriggers.kubeless.io                   2019-10-14T08:18:28Z
functions.kubeless.io                         2019-10-14T08:18:28Z
httptriggers.kubeless.io                      2019-10-14T08:18:28Z

开发函数，主要通过golang实现数组求和

package main

import (
    "github.com/kubeless/kubeless/pkg/functions"
    "strconv"
    "strings"
)

func Sum(event functions.Event, context functions.Context) (string, error) {
    strs := strings.Split(event.Data,",")
    var result = 0;
    for _,str := range strs {
        value,_ := strconv.Atoi(str)
        result += value
    }
    return strconv.Itoa(result), nil
}

发布function

• 通过kubeless发布

### flink：           表示function名称
### runtime：         表示运行时,实际对应镜像，支持的类型可通过kubeless get-server-config查看
###image-pull-policy：镜像拉去策略，默认Always，这里无法连接外网设置为IfNotPresent 
### from-file：       表示function源文件
### handler：         表示function具体函数名称，模块.函数名称
如果需要有依赖包通过dependencies指定，
--dependencies Gopkg.toml
[root@node4 kubeless]# kubeless function deploy flink \ 
>                             --runtime go1.12 \
>                             --image-pull-policy IfNotPresent \
>                             --from-file flink.go \
>                             --handler flink.Sum
INFO[0000] Deploying function...                        
INFO[0000] Function flink submitted for deployment      
INFO[0000] Check the deployment status executing 'kubeless function ls flink'

#查看发布的functions和对应的pod
[root@node4 kubeless]# kubeless function ls flink
NAME    NAMESPACE   HANDLER     RUNTIME DEPENDENCIES    STATUS   
flink   default     flink.Sum   go1.12                  1/1 READY
[root@node4 kubeless]# kubectl get pod
NAME                     READY   STATUS    RESTARTS   AGE
flink-69c84f8fd7-hrf2w   1/1     Running   0          16s
#错误排查
[root@node4 kubeless]# kubectl get pod -l function=flink -o yaml

• 通过apiserver发布

apiVersion: kubeless.io/v1beta1
kind: Function
metadata:
  name: testfunc
spec:
  deployment:  ### Definition as per v1beta2.Deployment
    metadata:
      annotations:
        "annotation-to-deploy": "final-value-in-deployment"
    spec:
      replicas: 2  ### Final deployment gets Replicas as 2
      template:
        metadata:
          annotations:
            "annotation-to-pod": "value"
  deps: ""
  function: |
    module.exports = {
      foo: function (req, res) {
            res.end('hello world updated!!!')
      }
    }
  function-content-type: text
  handler: hello.foo
  runtime: nodejs8
  service:
    ports:
    - name: http-function-port
      port: 8080
      protocol: TCP
      targetPort: 8080
    type: ClusterIP

函数调用 <除了主动请求，还支持事件和数据流的输入方式>

#kubeless客户端方式
[root@node4 kubeless]#  kubeless function call flink --data '1,3,5'
9
#API接口方式，通常是Ingress->Service，为外部提供接口
[root@node4 kubeless]# kubectl get svc
NAME         TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
flink        ClusterIP   172.20.49.159   <none>        8080/TCP   6m18s
kubernetes   ClusterIP   172.20.0.1      <none>        443/TCP    5d
[root@node4 kubeless]# curl -L --data '1,3,5,7,8' \
> --header "Content-Type:application/json" \
> 172.20.49.159:8080/api/v1/namespaces/default/services/flink:http-function-port/proxy/
24

调用参数模板

event:                                  
  data:                                         # Event data
    foo: "bar"                                  # The data is parsed as JSON when required
  event-id: "2ebb072eb24264f55b3fff"            # Event ID
  event-type: "application/json"                # Event content type
  event-time: "2009-11-10 23:00:00 +0000 UTC"   # Timestamp of the event source
  event-namespace: "kafkatriggers.kubeless.io"  # Event emitter
  extensions:                                   # Optional parameters
    request: ...                                # Reference to the request received 
    response: ...                               # Reference to the response to send 
                                                # (specific properties will depend on the function language)
context:
    function-name: "pubsub-nodejs"
    timeout: "180"
    runtime: "nodejs6"
    memory-limit: "128M"

5.kubeless-UI

#下载地址：https://github.com/kubeless/kubeless-ui/releases
#修改内部的imagePullPolicy策略
[root@node4 kubeless]# kubectl apply -f kubeless-ui.yaml
[root@node4 kubeless]# kubectl get svc -n kubeless
NAME   TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE
ui     NodePort   172.20.207.2   <none>        3000:30498/TCP   3h23m

访问测试

image.png