技术背景

• wsl2
• kind
• k8s
• k6 Operator(方案1)
• k8s worker deoployment（方案2）

  阿里云镜像仓库

  由于经常遇到time out问题，故在阿里云留个档。
  https://cr.console.aliyun.com/cn-guangzhou/instance/repositories

  $ docker login --username=dingtalk_yqsosa crpi-mstj0ugnrdkhzm1r.cn-guangzhou.personal.cr.aliyuncs.com
  $ docker tag [ImageId] crpi-mstj0ugnrdkhzm1r.cn-guangzhou.personal.cr.aliyuncs.com/chensuixaing/k6:[镜像版本号]
  $ docker push crpi-mstj0ugnrdkhzm1r.cn-guangzhou.personal.cr.aliyuncs.com/chensuixaing/k6:[镜像版本号]

多cluster部署

wsl2中尝试

• kind create cluster --name k6-kind-cluster --config kind-3nodes.yaml
• kind delete cluster –name k6-kind-cluster

  //尝试在wsl2中通过kind创建集群，结果 get node一直notready，最终放弃改换其他方式了
  kind: Cluster
  apiVersion: kind.x-k8s.io/v1alpha4
  nodes:
    - role: control-plane
    - role: worker
    - role: worker

  docker desktop中尝试

• 设置中可以直接开启，看日志也是pull这个镜像来实现的。需要梯子，最好手动pull。docker pull docker/desktop-cloud-provider-kind:v0.3.0-desktop.3
• 无法找到docker-desktop集群kubectl config use-context docker-desktop

  //解决方式
  1、确认docker desktop中node节点均正常
  2、设置-resources-wsl integration中必须启用 ubuntu
  3、临时测试 kubectl --kubeconfig /mnt/c/Users/12986/.kube/config get node
  4、echo 'export KUBECONFIG="/mnt/c/Users/12986/.kube/config"' >> ~/.bashrc
  5、source ~/.bashrc
  
  kubectl config get-contexts
  kubectl config use-context docker-desktop

  应用部署

  kubectl apply -f k6-worker-deployment.yaml
  kubectl apply -f k6-worker-service.yaml
  kubectl delete -f k6-worker-deployment.yaml
  kubectl delete -f k6-worker-service.yaml
  kubectl get deployments
  kubectl get services

  // k6-worker-deployment.yaml
  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: k6-workers
  spec:
    replicas: 3
    selector:
      matchLabels:
        app: k6-worker
    template:
      metadata:
        labels:
          app: k6-worker
      spec:
        containers:
        - name: k6-worker-container
          image: crpi-mstj0ugnrdkhzm1r.cn-guangzhou.personal.cr.aliyuncs.com/chensuixaing/k6:20251206
          command: ["./k6", "run", "dummy.js"]
          imagePullPolicy: Always
          env:
          - name: K6_WORKER
            value: "true"
          ports:
          - containerPort: 6565
            name: k6-dist-port

  // k6-worker-service.yaml
  apiVersion: v1
  kind: Service
  metadata:
    name: k6-worker-service
  spec:
    selector:
      app: k6-worker
    ports:
    - protocol: TCP
      port: 6565
      targetPort: 6565

  一些思考

1. 我想要手动执行压测，而不是apply之后直接执行

• k6 Operator似乎可以实现，但由于wsl2环境未能打通，暂未执行
• 实际上不一定要追求手动同时触发，多个pods同时执行，相差不了多少时间。
• k8s Job 或 CronJob

2. 日志的聚合

• 官网有相关https://grafana.com/docs/k6/latest/set-up/set-up-distributed-k6/usage/extensions/
• 应该单独起个influxdb，然后通过grafana进行聚合。dcokerfile中应当调整run命令，增加--out influxdb=http://<host>:<port>/<database>去尝试。有待实际行动。