Bcachefs is fucking awesome (conceptually) I hope it gets the love it needs.
is bcachefs stable?
just some day ago, someone posted this link:
https://kevincox.ca/2023/06/10/bcachefs-attempt/and that doesn’t sound great
anyone with more experience?
I only played around with Btrfs, which was quite nice, but I settled for a ZFS for my simple Raid 1’s, as I really liked it under BSD. else I just stick with Ext4.As mentioned in a child I definitely didn’t find it stable. It seemed to use masive amounts of CPU for unclear reasons and locked up a handful of times in my testing. When force-rebooting the filesystem was corrupted and seemed to continue to lose data after that point.
But that being said I think it may actually be good to merge it. It seems that there is lots of interest and the maintainers will be around to keep improving it. Getting it in the mainline (hopefully with some sort of “beta” label) will help it get the testing it needs. If the work is going to be done either way it seems best to be working in the main repo rather than keeping it separate and needing to continue handling migrations and merge conflicts. Look at BTRFS. It was known for data loss but now seems to be pretty stable with lots of eyes and lots of work.
But that being said I think it may actually be good to merge it. It seems that there is lots of interest and the maintainers will be around to keep improving it.
Yeah I think people shouldn’t hold it against bcachefs to have some issues in experimental stages and going mainline is a good way to catch obscure & rare bugs.
Look at BTRFS. It was known for data loss but now seems to be pretty stable with lots of eyes and lots of work.
IMHO it’s pretty unfair how people like to give new, complex, filesystems a ‘reputation’ immediately, when there are some issues. I hope not the same is done with bcachefs and it gets its fair shake. Occasional issues popping up now (like in your blog post), hopefully, will also allow some of its cult followers to touch grass and get a reality check (filesystem = difficult). IMHO Kent really should remove the obnoxious “The COW filesystem for Linux that won’t eat your data.”-sentence from his website as it encourages such nonconstructive attitudes. I’m sure he is aware that, at this point, btrfs is less likely to eat your data by many orders of magnitude compared to his draft filesystem (and that’s mainly because most of those data eating bugs have been found and fixed in btrfs, not because it’s somehow impossible to corrupt by design).
