Developing Microservices Frictionlessly

My sprint talk for London Microservices User Group on 2 November 2016.

https://www.meetup.com/London-Microservices-User-Group/events/234613236/

Transcript

1. DEVELOPING MICROSERVICES FRICTIONLESSLY London Microservices User Group 2 Nov 2016
2. Paris Kasidiaris Co-founder at I ❤ Python, containers and pragmatic programming.  @pariskasid lair.io
3. A look on our stack web: Django / Celery auth: Flask event_watcher: Python proxy: NGINX / Ceryx db: Postgres cache: Redis
4. How we build microservices We do tests/CI/CD on Docker We run production on Docker We orchestrate with Docker Compose
5. But, how do we develop microservices? We took a look at a few options
6. Run Docker and the whole stack in a VM Ditched Slow to download images and production data Host ↔ VM ↔ container volumes are a mess Not easy to get feedback (we are a remote-first team)
7. Develop locally and mock services Ditched Does not run the whole stack ➡ wild bugs appeared Slow to download images and production data Not easy to get feedback (we are a remote-first team)
8. We layed down the requirements to fix this Fast and easy to build dev environments Containers should be accessible via the internet Working offline should be an option
9. Solution Set up a dev server containing all Docker images Connect local Docker client to dev server's Docker sync.sh: Sync local workspace with dev server via Unison watch.py: Expose containers via NGINX a er docker start
10. Takeaways Fast and disposable environments Resource saving (common build cache and images) Accessible containers via a URL Offline-fallback since we use Docker Compose
11. Technical challenges Authentication of multiple users with one Docker server Synchronization of local files to the proper remote directory Dynamic allocation of URLs for each container port Thanks ! sourcelair/ceryx
12. Eventually We bundled all these in a tool named . Stolos
13. Thank you. Check out what we built at ! stolos.io