Hacker News Reader

Redpanda is built on top of Seastar, as is ScyllaDB, so the reasons might be closely related: https://www.scylladb.com/2016/02/09/qualifying-filesystems/

I was so happy about just running a single small binary for kafka, but then I saw this requirement that is harder to fulfill :/

Why, you can run on a single core and 1G of ram.

We just give you optimal prod settings for saturating boxes.

See rpk container

So I don’t need XFS to run a single-node Redpanda cluster?

correct. xfs is just what we test all the time.

Mostly for sparse file system support. We dispatch IO eagerly and out of order.

How much of a hard requirement is this? Can I play around on non XFS?

This is not a hard requirement, but XFS is highly recommended for production.