Hacker News new | past | comments | ask | show | jobs | submit login
Lessons Learned: Scaling Uber to 2000 Engineers, 1000 Services, and 8000 Git Repos (highscalability.com)
28 points by riqbal on Oct 12, 2016 | hide | past | favorite | 7 comments



I'm most curious how 1000 services fit in 8000 repos.

(Maybe it's the same way that 40 cities fit in 70 countries.)

---

FYI, with Git, the "repo too big" problem has a lot of solutions.

Grafts (shallow commits) for long histories, large file storage (LFS) for binaries, and in extreme cases filter-branch.

Yes, when you hit Google scale is it possible to write enough source code to not fit on disk. But Uber isn't Google scale. Maybe Uber can't fit code into one repo. But ten repos should be fine.


The 40 is probably a typo, the cities number is closer to 400.

I don't think the microservices is about repo too big in a technical sense. The problem is when you have 100s of engineers working on a single repo, just merging changes or building , testing etc. will take all day. Microservices are more a solution to scaling engineering teams with a lot of people because they can work independently rather than a technical scaling thing.


Sure, but I wasn't quibbling about the number of microservices, just about the number of repos used.


"This same pattern of explosive growth is happening in cities all over the world. In fact, Uber is now in 40 cities and 70 countries."

Um. Stopped reading at this point.


even worse, they repeat it further down in a bullet point list


Not in the article: why does Uber need that much software? Does anyone know here what all those services and engineers do?


Wow Uber'S tech stack is a total mess




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: