Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

If you're pushing the database up into the application layer, do you have to route all write operations through a single "master" application instance? If not, is there some multi-master scheme, and if so, is it cheaper to propagate state all the time than it is to have the application write to a master database instance over a network? Moreover, how does it affect the operations of your application? Are you still as comfortable bouncing an application instance as you would otherwise be?


The answer is: yes, you do have to write through a single primary application instance.

So far.

The two important things here are:

1. Fly.io makes it really easy to write through a single primary application instance

2. There are ways to solve this problem so your application doesn't have to worry about it.

Right now, you have to be a little careful bouncing app instances. If you bounce the writer, you can't perform writes for 15s or whatever. This is a big problem during deploys.

There are a tremendous number of Fly.io users that are fine with this limitation, though. It's pretty valuable for some segment of our customers right now.


It definitely seems like it could be useful for some use cases; I'm just trying to get my head around the constraints. :)


What are some ways alluded to in number 2?




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

Search: