I keep being tempted by the flexibility of the mirror-layouts of btrfs and bcachefs, I have so many old disks that are smaller than my main disks but still a useful amount of space. But ZFS is just so mature and reliable, and the newer contenders seem to still fight with such serious bugs… it’s very hard to convince myself to jump over.
The ZFS ecosystem also has some really mature snapshot-based backup system that both snapshot the local disk and can do send/recv to copy to backup disks locally or remotely… and clean up old snapshots. So I’d be signing up to replace that as well.
Interesting read. My understanding is that BTRFS has had long standing issues with it’s parity raid configurations (I.e. Raid5/6). Recently, kernel version 6.2 seems to have added more fixes. I’ll stick to the “wait and see” approach for the time being, as Raid1 is sufficient for my home server.
Other than the Raid5/6 issue, I don’t know of any other issues with BTRFS losing data. If there are some, I’d love to know.
deleted by creator
I don’t really know how it happened
Better hope it was cosmic radiation and not an early indicator of failing hardware.
deleted by creator
Reminds me of a time I ended up getting some corruption in some cache file for firefox. It was after I did a number of forced power offs. I’m not sure if I should blame such corruption on the filesysten.
deleted by creator
I think the upcoming on-disk format changes are going to fix most of the RAID5/6 issues.
RAID5/6 are clearly marked as unstable in the docs, so that doesn’t bother me too much. But there is definitely a history of corruption. I can’t actually find any concrete info these days as the act seems to have been cleaned up recently. For example Red Hat started supporting BTRFS but dropped official support: https://archive.is/1VuTK
I think it will eventually be the best alternative, but I won’t be using bcachefs for the following years… for anything server or professional related needs, I’d go with ZFS. In my personal systems, I use BTRFS including on NixOS.
Yeah, that is exactly my thoughts right now. The API and features of bcachefs are excellently designed. I can’t comment on the implementation other than the bugs I have seen. But I really hope that the bugs get squashed out because I would love to have those features.
Interesting read on bcachefs. Thanks for that. I have been keeping an eye on it for a while now but didn’t dare to try it out.
Very unfortunate to read. I’ve wanted to try bcachefs for a while on my PC (right now I’m using bcache with ext4 on top) but maybe it’s not the time yet. Especially since my PC is already prone to random lockups (thanks, AMD.)
Well that’s not reassuring, but I do wonder how prevalent data lose is with bcachefs. It has all the features I want but have been holding off due to not wanting to run a different kernel, or add complexity to my system.