onatm

I haven’t put too much thought into this yet. A "branch" depth threshold is the first thing that comes to mind, with RBD flattening as an option once that threshold is reached.

The performance definitely sucks but it's not a really serious project. I wouldn't use something like Homescale for using facing products.

It was a good read. I also worked on a similar product that used zfs instead of Ceph.

Also, as you said, Homescale is a lot closer to Xata. It all started as a joke and the name stuck.

I am not sure where you think I claimed it'll be "easy and quick". Do you really think building a system on top of k8s internals and Ceph is something that everybody can pull off?.

It's far from complete. This is just the infrastructure bit that I need to build the actual service. I am going to build an operator, a coordinator, an API and a CLI. I am going to continue writing posts while building them.

I hope they'll enjoy reading it.

Hey HN, I vibe-coded a small self-hosted phonics flashcard app for my daughter this afternoon. She has an "exam" next week based on the words in the data folder. If your kids struggle with phonics, you can use this too: generate word lists with any LLM, drop them into the data folder, and follow the README.