Menu
Back to Discussions

So, when do you really need a distributed transaction?

I've been thinking a lot about the extra work and hassle that comes with distributed transactions, especially when you're dealing with microservices. It just feels like every time we even think about using one, we're piling on a bunch of new stuff just to keep everything straight and deal with things breaking. I get that there are times when you absolutely need everything to be super consistent across different services, like with money stuff. But do we sometimes jump to them when something simpler, like eventual consistency or a saga, would work just fine and be way less complicated? So, what are the big signs or must-have requirements that really make you say, "Okay, we need a distributed transaction here"?
4 comments

Comments

Loading comments...