我們與Kubernetes的距離

1. Kubernetes簡介與近況更新
2. 容器部署工具(Helm)簡介
3. K8s如何加速開發流程與服務部署(實戰與案例分享)
4. K8s最佳實踐

Transcript

1. 我們與 Kubernetes 的距離 Astley Chen GDG Cloud Taipei Co-Organizer Cloud Solution Architect 2019/12/12
2. Who AM I Job Titles: • Software Developer • DevOps/SRE Engineer • Cloud Solution Architect • Data Engineer Enthusiasms : • Cloud Native Computing / Docker • Kubernetes / Helm • Service Mesh / Istio • Infrastructure Managing 8 clusters(5 on GKEs, 3 on EKS) more than 80 releases.
3. Agenda 1. Kubernetes簡介與近況更新 2. 容器部署工具(Helm)簡介 3. K8s如何加速開發流程與服務部署(實戰與案例分享) 4. K8s最佳實踐
4. 什麼是Kubernetes? • Kubernetes is an open source system for managing containerized applications across multiple hosts. • Open-sourced by Google in 2015, donated to CNCF. Kubernetes簡介與近況更新 3/25 1.14.0 6/17 9/18 12/9 1.15.0 1.16.0 1.17.0
5. Kubernetes的優勢與價值 • Automate containers deploying, scaling, load balancing • Automatic rolling updates and rollbacks • Automatically handle networking, storage, logs, alerting • Health check & Self-healing • RBAC advanced security control • Reducing resource costs • Ease-of-use and portability • Outstanding community and industry support ◦ GKE - Google Kubernetes Engine ◦ EKS - AWS Elastic Kubernetes Service ◦ AKS - Azure Kubernetes Service ◦ ... Kubernetes簡介與近況更新
6. Managed Kubernetes Service Kubernetes簡介與近況更新
7. Kubernetes 架構 Kubernetes簡介與近況更新
8. Kubernetes Deployment Kubernetes簡介與近況更新 Deployment: ★ Create ★ Scale ★ Rollout ★ Rollback ★ Pause POD Pod: ★ Containers ★ Network ★ Volumes ★ Resources ★ PullSecrets ★ PullPolicy ★ ...
9. Kubernetes API Resources Kubernetes簡介與近況更新 • WORKLOADS ◦ Container v1 core ◦ CronJob v1beta1 batch ◦ DaemonSet v1 apps ◦ Deployment v1 apps ◦ Job v1 batch ◦ Pod v1 core ◦ ReplicaSet v1 apps ◦ ReplicationController v1 core ◦ StatefulSet v1 apps • DISCOVERY & LOAD BALANCING ◦ Endpoints v1 core ◦ Ingress v1beta1 extensions ◦ Service v1 core • CONFIG & STORAGE ◦ ConfigMap v1 core ◦ Secret v1 core ◦ PersistentVolumeClaim v1 core ◦ StorageClass v1 storage.k8s.io ◦ Volume v1 core
10. Helm • Chart ◦ 定義要被執行的應用程式中，所需要的工具、資源、服務等資訊 • Release ◦ 一個執行於 Kubernetes 的 Chart 實例 • Repository ◦ 主要用來存放 Chart 的倉庫，如 KubeApps、Bitnami。 ◦ Public repos ▪ stable https://kubernetes-charts.storage.googleapis.com ▪ incubator https://kubernetes-charts-incubator.storage.googleapis.com/ ▪ gitlab https://charts.gitlab.io ▪ bitnami https://charts.bitnami.com Helm is a Kubernetes package and operations manager https://k2r2bai.com/2017/03/25/kubernetes/helm/quick-start/
11. Helm Components • Helm Client ◦ Client-side CLI ◦ Communicate with Tiller Server ◦ Charts release CRUD operations ◦ Charts release version control • Tiller Server ◦ Receive commands from client ◦ Communicate with API-Server ◦ Manage releases and API resources ◦ Removed from Helm v3 (released 11/13) 容器部署工具簡介 https://k2r2bai.com/2017/03/25/kubernetes/helm/quick-start/
12. Helm • helm create • helm template • helm install/upgrade • Helm rollback • helm delete 容器部署工具簡介
13. 實戰與案例分享 - 推薦引擎 K8s如何加速開發流程與服務部署 [D] Data Source and Data Preparator [A] Algorithm [S] Serving [E] Evaluation Metrics http://predictionio.apache.org/customize/
14. 需求列表 • DevOps ◦ GitLab.com (K8s integration) ◦ Runners for CICD tasks • Kubernetes Cluster ◦ Dedicated Node Pools ◦ Pods-Nodes Affinity ◦ Horizontal Pod Autoscaler ◦ Statefulsets Operation ◦ Service Mesh • Workflow Management • Messaging Services • Monitoring and alerting 實戰與案例分享 - 推薦引擎
15. 需求列表 • Kubernetes Engine ◦ Running all microservices and database clusters to fulfill business logic and scenarios. ◦ DevOps & CI/CD by integration with GitLab • PubSub & Cloud Storage ◦ Streaming log and batch data ETL ◦ Blob objects repository for backup and restore. • BigQuery & DataStudio ◦ Data warehouse for streaming log and batch data ◦ Data analytics and presentation • G-Suite & OAuth API ◦ IAM management & apps access control • Stackdriver ◦ Infrastructure & Kubernetes monitoring and alerting 實戰與案例分享 - 推薦引擎
16. DevOps 實戰與案例分享 - 推薦引擎 .gitlab-ci.yml
17. Kubernetes Cluster • Cluster Autoscaler • Node Taints and Pod Tolerations • Node Labeling and Node Selector • Node Affinity • Pod Autoscaler 實戰與案例分享 - 推薦引擎
18. Kubernetes Cluster • Statefulset Operation ◦ Elasticsearch cluster ◦ MongoDB Replica Set ◦ MariaDB cluster • Challenges ◦ Upgrade statefulset spec ◦ Upgrade PVC ◦ Update strategies ▪ On Delete ▪ Rolling Updates 實戰與案例分享 - 推薦引擎
19. Workflow管理系統 - Airflow • DAG • Celery Executor • KubernetesPodOperator 實戰與案例分享 - 推薦引擎
20. Charts As Building Blocks 實戰與案例分享 - 推薦引擎
21. K8s最佳實踐 • From Ground Up ( Docker → Docker Compose → Helm ) • Configure Liveness, Readiness and Startup Probes Carefully • Configure Resource Requests & Limits Honestly • Service Mesh(Istio) is MUST for Traffic Routing and Tracing • Internal Subsystems Health Probes
22. K8s最佳實踐 - Bonus #蔡宗城 - 聽說未來不會 K8S 的話，將會被淘汰？ ★ Google Cloud Platform：US$ 300 頭12個月免費額度，可以使用來 開 GKE ▍連結：https://cloud.google.com/free/ ★ Microsoft Azure：US$ 200 頭12個月免費額度，而且 AKS 有關 AI & ML 的 Workload 是免費的 ▍連結：https://azure.microsoft.com/en-us/free/search/ ★ IBM Cloud：免費提供單一 Worker Node 的 K8S 叢集，對於初學者 來說算是滿夠了 ▍連結：https://www.ibm.com/cloud/free/ ★ Alibaba Cloud：US$ 300 頭12個月免費額度，對於某些特定的 K8S 服務永久免費 ▍連結：https://www.alibabacloud.com/campaign/free-trial ★ Redhat OpenShift：60 天的單一 Worker Node K8S 叢集，可以在 Redhat Account 中取得 ▍連結：https://www.openshift.com/learn/get-started/ ★ Katacoda：把玩 K8S 最常被使用的平台之一，有各種不同類型的 K8S 叢集可以使用，甚至有 minikube 版本 ▍連結：https://www.katacoda.com/courses/kubernetes ★ Tryk8s：提供免費嘗試 K8S 的環境 ▍連結：https://tryk8s.com/ ★ CodeFresh：提供 US$ 500 的免費額度讓使用者嘗試使用 K8S (背 後是利用 GCP 完成) ▍連結：https://codefresh.io/google-cloud/ ★ KubeSail：使用 GitHub 帳號註冊就可以獲得一個免費的 K8S 叢集 ▍連結：https://kubesail.com/deployments https://medium.com/techprimers/free-tiers-in-different-cloud-platforms-for-trying-out-kubernetes-2ccda3f296dc
23. 參考資料 Kubernetes Basics • https://kubernetes.io/docs/tutorials/kubernetes-basics/ • https://blog.newrelic.com/engineering/what-is-kubernetes/ • http://omerio.com/2015/12/18/learn-the-kubernetes-key-concepts-in-10-minutes/ • https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/ Helm Basics • https://k2r2bai.com/2017/03/25/kubernetes/helm/quick-start/