> zfs on Linux has not been production ready for decades. People have lost data from it.
I don't think that's true. Other than with ZFS-native encryption, which I grant has been less reliable, it's been rock solid for a very long time. And I've run >1PB of postgres databases on it professionally, so I feel fairly comfortable in that assertion.
> There's no real reason to allow the default installer to do this.
The default Ubuntu installer at least used to support ZFS, which is the point.
If you Google zfs Linux data loss, you can find many posts about this. Including one lengthy discussion on HN.
Also, you are not the typical user installing the OS from the default installer. I am not saying ZFS is bad, but not including it in the default installer is no big deal.
So funny thing. I was planning to agree and write something about how you can find data loss stories for literally any filesystem, but the relative frequency and nature of those stories is important to differentiate.
https://duckduckgo.com/?t=ffab&q=zfs+data+loss&ia=web - First result is someone claiming data loss after setting sync=disabled on a single disk over usb with slog+l2arc on the same disk, and scrub turns up 50k errors; i.e. they did everything they could to hold it wrong, and in the end their disk physically failed, which I really don't think is a ZFS problem. Second result is a stack overflow thread discussing why ZFS doesn't fail the usual ways. Third result is official docs. If I go down the rest of the first page of duckduckgo results, it's mostly discussions of how ZFS protects against data loss, with the one exception of https://forum.level1techs.com/t/solved-zfs-monthlong-changes... ... which looks bad until you realize the use didn't mount a filesystem, and once they found it they recovered their data just fine.
So no, based on random web searches I conclude that ZFS remains head and shoulders above every other option.
---
Edit: If I search for "zfs Linux data loss hacker news", I get https://news.ycombinator.com/item?id=22005181 which appears to contain zero stories of losing data on ZFS, although there's a bunch of stories about BTRFS doing so. Most of the remaining results are news stories about a single bug from 2023 and one story about ZFS's native encryption having problems (which I grant is a footgun).
I don't think that's true. Other than with ZFS-native encryption, which I grant has been less reliable, it's been rock solid for a very long time. And I've run >1PB of postgres databases on it professionally, so I feel fairly comfortable in that assertion.
> There's no real reason to allow the default installer to do this.
The default Ubuntu installer at least used to support ZFS, which is the point.