Progettare un ambiente di sviluppo locale basato sui container

Un ambiente di sviluppo locale efficiente facilita l’avvicendamento degli sviluppatori su un progetto, aiuta la condivisione delle informazioni e mette gli sviluppatori nella condizione di poter eseguire il build in locale in un ambiente simile a quello di produzione. Nel caso di ambienti di produzione basati su container la scelta più logica è svolgere l’attività di sviluppo software all’interno di container. Tuttavia, la progettazione di un ambiente di sviluppo locale efficiente pone molte sfide. In questo intervento presenterò un approccio efficace per l’implementazione di un ambiente di sviluppo locale di facile manutenzione che permetta iterazioni veloci, rispetti le preferenze degli sviluppatori e non disabiliti meccanismi specifici dello stack applicativo (ad esempio hot-reload).

Transcript

1. Progettare un ambiente di sviluppo locale basato sui container Martino Fornasa @mfornasa - www.fornasa.it - www.kiratech.it
2. • Fast developer onboarding • Software dependency management • Production-like environment • Local testing • Build knowledge Local Development Workflow @mfornasa
3. Creating a local development environment is hard @mfornasa
4. Creating a local development environment is hard and very language-specific @mfornasa
5. @mfornasa # Dockerfile 
 ... ADD package.json yarn.lock /cache/ ADD .yarn-cache.tgz / RUN cd /cache \ && yarn \ && cd /opt/app && ln -s /cache/node_modules node_modules \ && tar czf /.yarn-cache.tgz /usr/local/share/.cache/yarn COPY . /opt/app Node.js Dockerfile Tricks
6. So I rolled my own solution @mfornasa # docker-compose.yml services: app: build: app/. command: yarn run start-dev depends_on: - db ports: - '8080:8080' volumes: - ./app/.:/opt/app db: image: postgres restart: always environment: ... ports: - '5432:5432' # Dockerfile FROM some-node-on-alpine COPY . /opt/app RUN yarn CMD ["yarn", "run", “watch"] # index.js var express = require('express'); var app = express(); var server = app.listen(3000,function() { console.log("Listening on port 3000"); });
7. What could have possibly gone wrong? @mfornasa
8. @mfornasa Speed Sync Issues Cleanup
9. Resources @mfornasa
10. People hated it… @mfornasa
11. • Reduce friction • Make it easy to do the right thing • Allow people to use their preferred tools (IDE, linters, debuggers, etc.) • Low latency! Implementation goals @mfornasa
12. A development workflow for Docker Kubernetes applications @mfornasa
13. • Shift left the environmental parity • Make it easier to setup complex environments (databases, cache engines, search engines) • Microservices • Same (or similar) tools for dev and deploy • Each developer being able to run the full system (maybe also connecting to external systems) • CI tool builds the final image and enforce policies Benefits of a Kubernetes workflow @mfornasa
14. 1. Auto rebuild and deploy 2. Run on cluster + Sync filesystem
 3. Run locally + Bridge network from: Joe Beda, TGI Kubernetes 039 @mfornasa Three approaches
15. • Higher guarantee of correctness • Easy to do on a remote cluster • Better choice for compiled languages • Speed issues Auto rebuild and deploy @mfornasa 1.
16. @mfornasa https://draft.sh/ • Build, push and deploy • Detect language, generates Dockerfile + Helm Chart • Supports out of the box: .NET, Go, Node, PHP, Java, Python, Ruby • In progress: • Full-fledged remote debugging support • Improved support for multiple components draft init draft create draft up draft connect
17. @mfornasa • Watches your source code and the dependencies of your Docker images for changes and runs a build and deploy when changes are detected • Streams logs from deployed containers • Continuous build-deploy loop, only warn on errors
 https://github.com/GoogleContainerTools/skaffold
18. @mfornasa https://github.com/GoogleContainerTools/skaffold apiVersion: skaffold/v1alpha4 kind: Config build: artifacts: - image: gcr.io/k8s-skaffold/node-example context: backend sync: '*.js': . deploy: kubectl: manifests: - “k8s/**"
19. • Higher chance of drifting from the image built by CI • Better choice for interpreted languages • Usually faster Run on cluster Sync filesystem @mfornasa 2.
20. @mfornasa • Sync files with the container • Server component (DaemonSet) • In progress: remote command execution https://github.com/vapor-ware/ksync ksync init ksync watch & kubectl apply -f config.yaml ksync create --selector=app=app $ (pwd)/ksync /code ksync get
21. • Running code locally • Developing against remote resources Run locally Bridge network @mfornasa 3.
22. @mfornasa https://www.telepresence.io/
23. We are hiring! @mfornasa https://www.kiratech.it/work-with-us/
24. • Jez Humble, David Farley, Continuous Delivery, ch. 3. • Joe Beda, TGI Kubernetes 039 • Daniel Bryant, Creating an Effective Developer Experience on Kubernetes References @mfornasa