#pod存活性探测 exec: #使用exec，在容器里执行shell命令，返回0则成功 command: ["test", "-e", "/tmp/xxhealthy"] #执行的命令 # kubectl get pod #查看pod， deployment dep-xxx --type=NodePort --name=svc-test1 #将 deploy里的端口暴露出来 # kubectl expose pod pod-nginx-web1 --type=NodePort --name=svc-test2 #将 pod里的端口暴露出来 # kubectl get svc NAME ★创建PV（使用RBD后端） # cat > pv-rbd-image-test1.yaml <test1 #此PV的名称 labels: pvlab: pvlab-rbd-image-test1 spec: capacity:

k8s-app-name app-name ai-test ai-dc-server ai-dc-server ai-dc-servedr ai-preview ai-dc-web ai-dc-web ai-dc-web ai-prod ai-dc-api ai-dc-api ai-dc-api 业务线名称 ai dt ad 现有环境名 test preview prod 统一规划环境名和业务应用名，适配标准自动化运维。 k8s-namespaces 环境名称定义采用业务线缩写名加环境名组成 k8s-service名称、app名称和应用名称包名保持一致 k8s-api配置对象 作用 k8s-namespace 通过配置文件关键字dev/test/prod等声明应用所属的环境，隔离不同环境业务，通过特定标识来识别业务线。 k8s-service k8s-dns注册服务名，通过配置文件关键字关联业务线应用名称，保持应用和k8s之间的关联。 仓库域名+路径 空间名 应用名称 日期-时间戳 git版本库 镜像完整地址 registry.hz.local/huize ai-test ai-dc-web 20190510-1033 v20 registry.hz.local/huize/ai-test_ai-dc-web:20190510-1033_v20 k8s镜像构建过程 domain/path namespaces app-name

⾼级活动探针示例 活动探测由kubelet执⾏，因此所有请求都会在kubelet⽹络命名空间中进⾏。 apiVersion: v1 kind: Pod metadata: labels: test: liveness name: liveness-http spec: containers: - args: - /server image: gcr. ），你应该附加标识语义的Label属性，例如 { app: myapp, tier: frontend, phase: test, deployment: v3 } 。 这将允许您选择适⽤于上下⽂的对象组——例如，所有标记了 tier: frontend 的Pod的 Service；或“myapp”应⽤的所有“test”阶段的组件。有关此⽅法的示例，请参阅 guestbook 应⽤程序。 可通过简单的、从其选择器 kubectl describe nodes 命令检查Node的容量和数量。 例如： $ kubectl describe nodes e2e-test-minion-group-4lw4 Name: e2e-test-minion-group-4lw4 [ ... lines removed for clarity ...] Capacity: alpha.kubernetes

VM VM VM VM VM VM VM VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Test VM K8S Test VM K8S Test VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM VM VM VM VM VM VM VM K8S Prod VM K8S Prod V K8S Prod VM K8S Prod VM K8S Test VM K8S Test VM K8S Test VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod (Master) (Master)

VM VM VM VM VM VM VM VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Test VM K8S Test VM K8S Test VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM VM VM VM VM VM VM VM K8S Prod VM K8S Prod V K8S Prod VM K8S Prod VM K8S Test VM K8S Test VM K8S Test VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod VM K8S Prod (Master) (Master)

Performance of a cluster in different time slot may differ • Due to CPU oversold • Suggestion: • Run the test against the same cluster during near time • Make CPU the bottleneck • 1 CPU handles 500,000 pps measurement Test topology Test result Service type Short connection cps Short connection P99 latency Long connection pps ClusterIP +40% -31% not available NodePort +64% -47% +22% Test result • Perf

CI and CD Local development Gitlab runner Dockerhub Pod Pod Pod -Build -Test -Push to dockerhub -Deploy Application infrastructure The services around Kubernetes Take control autodiscovery Grafana dashboards Graylogs search and dashboards Performance test Performancetest: LOCUST ● Test as code ● Highly scalable ● Less hardware consuming Connections LOCUST: Fast

靜態應用安全測試 (白箱測試): Static Application Security Test (SAST) 靜態應用安全測試 : Interactive Application Security Test (IAST) 動態應用安全測試 (黑箱測試): Dynamic Application Security Test (DAST) 軟體元件分析: Software Component Analysis

CI/CD Step1 Branch staging change Condition Action Object Build UT test Staging Job Deploy Staging Job IT Test Staing Job Merge Preview Branch Step2 Deploy Preview Job Merge

repeatable ML experiments with containers • Deploy ML components to Kubernetes with Kubeflow • Scale and test ML experiments with Helm • Manage training jobs and pipelines with Argo • Serve trained models for • Tensorflow Serving • Seldon Demo: Run TensorFlow Training with Kubeflow Demo: Scale and Test Experiments in Parallel using Kubernetes, TFJob, and Helm • Spin up pods for each variation of