We published a blog post on our journey from in-house Flink autoscaler to the OSS community based Flink autoscaler at Netflix: netflixtechblog.com/a-tale-of-two-flink-autoscalers…

This is a great writeup ❤️ We're on a similar journey at Yelp (say hi to Francesco for me 😉 )

@Mingliang Liu @Toby C we need more "war stories" like these .. Keep sharing your great work 🙌🙏

We've got some blog posts in the works 🙂

Amazing read, thanks for sharing! I would also really enjoy more. Personally, I’m very selfishly interested in how you all are managing this given your running large stateful jobs. This is a core issue I’m constantly struggling with 🙈, even on Flink 2.2.

@Francis Altomare if you are running into large state issues you might want to look into Apache Fluss 😄

Yeah, this is definitely something we need to do. Back when I was architecting our upgrades it was too new for me to get org-wide buy in. However, now it’s really cool to see the momentum — I’m really looking forward to getting the chance to play with it 👍

Most of our Flink jobs are still on 1.x. Yes we took a look at Fluss and will evaluate that for our use cases as well.

@Mingliang Liu anything you need feel free to reach out on Fluss slack and our community will be happy to help 🙏