自建cloudflare workerd边缘函数运行环境（k8s部署）

目录结构

root@k8s-master1:~/workerd# tree
.
├── Dockerfile
├── config.capnp
├── deployment.yaml
├── index.js
└── workerd  # 可以在官方github开源地址下载编译好的版本，https://github.com/cloudflare/workerd

1 directory, 5 files
root@k8s-master1:~/workerd#

workerd 本质上是一个 js 代码运行环境，而且因为其无状态的特性，非常适合放到 k8s 里。

我去找过 Cloudflare workerd 开源项目主页和 docker hub 官方的仓库，没看到有构建好的 docker 镜像，遂只能自己写 Dockerfile 构建，也很简单。

Dockerfile

# 选用轻量的 Debian 12 (bookworm) 镜像
FROM debian:bookworm-slim AS runtime

# 安装 TLS 根证书
RUN apt-get update && apt-get install -y --no-install-recommends \
    ca-certificates \
    tzdata \
    && rm -rf /var/lib/apt/lists/*

# 设置工作目录
WORKDIR /app

# 将 workerd 二进制文件和配置文件、代码复制进镜像
COPY workerd index.js config.capnp /app/

# 暴露 workerd 监听的端口（按 config.capnp 配置决定）
EXPOSE 8080

# 运行 workerd
ENTRYPOINT ["/app/workerd", "serve", "/app/config.capnp"]

workerd的配置文件config.capnp

using Workerd = import "/workerd/workerd.capnp";

const config :Workerd.Config = (
  services = [
    (name = "main", worker = .mainWorker),
  ],
  sockets = [
    ( name = "http", address = "*:8080", http = (), service = "main" )
  ]
);

const mainWorker :Workerd.Worker = (
  modules = [
    (name = "index.js", esModule = embed "index.js")
  ],
  compatibilityDate = "2024-01-01",
);

export default {
  async fetch(request, env, ctx) {
    return new Response("Hello from Cloudflare workerd on K8s! version:1.0.2\n");
  }
};

k8s声明配置文件deployment.yaml，其中ingress也在里面定义了。

apiVersion: v1
kind: Namespace
metadata:
  name: workerd
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: workerd-deployment
  namespace: workerd
spec:
  replicas: 10 # 副本数量
  selector:
    matchLabels:
      app: workerd
  template:
    metadata:
      labels:
        app: workerd
    spec:
      containers:
      - name: workerd
        image: workerd:v1.0.2 # 镜像名字
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 8080
---
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: workerd-ingress
  namespace: workerd
  annotations:
    # 允许的跨域配置或请求头大小等可在此扩展
    nginx.ingress.kubernetes.io/proxy-read-timeout: "60"
spec:
  ingressClassName: nginx
  rules:
  - host: workerd.k8s.cn # 域名
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service:
            name: workerd-service
            port:
              number: 8080
---
apiVersion: v1
kind: Service
metadata:
  name: workerd-service
  namespace: workerd
spec:
  selector:
    app: workerd
  ports:
  - name: http
    port: 8080
    targetPort: 8080

可执行程序 workerd 则可以在官方 github 开源地址下载编译好的版本，https://github.com/cloudflare/workerd

可以在其他主机构建好workerd的镜像并导出

docker build -t workerd:v1.0.2 .

docker save -o workerd-v1.0.2.tar workerd:v1.0.2

然后把 workerd-v1.0.2.tar 上传到 k8s 上的各个节点上并加载镜像到 k8s 的镜像命名空间，如果有镜像仓库的话也可以直接推送镜像上去，省略

ctr -n=k8s.io images import workerd-v1.0.2.tar

再 kubectl apply -f deployment.yaml 就能直接部署

检查 pod 和 svc、ingress

root@k8s-master1:~# kubectl get pod -n workerd -o wide
NAME                                  READY   STATUS    RESTARTS   AGE   IP             NODE        NOMINATED NODE   READINESS GATES
workerd-deployment-64644bdbc9-5gdsr   1/1     Running   0          4d    10.244.5.102   k8s-node3   <none>           <none>
workerd-deployment-64644bdbc9-fkg99   1/1     Running   0          4d    10.244.4.93    k8s-node2   <none>           <none>
workerd-deployment-64644bdbc9-h4zdg   1/1     Running   0          4d    10.244.7.94    k8s-node4   <none>           <none>
workerd-deployment-64644bdbc9-h6ncm   1/1     Running   0          4d    10.244.4.94    k8s-node2   <none>           <none>
workerd-deployment-64644bdbc9-njq8b   1/1     Running   0          4d    10.244.3.72    k8s-node1   <none>           <none>
workerd-deployment-64644bdbc9-pk8h7   1/1     Running   0          4d    10.244.5.101   k8s-node3   <none>           <none>
workerd-deployment-64644bdbc9-r2xzd   1/1     Running   0          4d    10.244.6.62    k8s-node5   <none>           <none>
workerd-deployment-64644bdbc9-vd99f   1/1     Running   0          4d    10.244.3.71    k8s-node1   <none>           <none>
workerd-deployment-64644bdbc9-wvs8z   1/1     Running   0          4d    10.244.6.63    k8s-node5   <none>           <none>
workerd-deployment-64644bdbc9-xfs4l   1/1     Running   0          4d    10.244.7.95    k8s-node4   <none>           <none>
root@k8s-master1:~#
root@k8s-master1:~# kubectl get svc -n workerd
NAME              TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)    AGE
workerd-service   ClusterIP   10.103.32.102   <none>        8080/TCP   4d
root@k8s-master1:~#
root@k8s-master1:~# kubectl get ingress -n workerd -o wide
NAME              CLASS   HOSTS            ADDRESS       PORTS     AGE
workerd-ingress   nginx   workerd.k8s.cn   10.10.0.200   80, 443   4d
root@k8s-master1:~#

root@k8s-master1:~# root@k8s-master1:~# curl --resolve workerd.k8s.cn:80:10.10.0.200 https://workerd.k8s.cn/
Hello from Cloudflare workerd on K8s! version:1.0.2
root@k8s-master1:~#

Cloudflare 虽然开源了 workerd ，但 workerd 本质上只是一个运行环境，并没有集成 KV 键值对存储和 D1 数据库，只能依赖外部服务来实现，调度、多租户引擎也没有开源。

如果想要更完善的且兼容 Cloudflare workerd 的边缘函数，可以看看基于 workerd 开发的 celld 项目

https://github.com/denoland/celld