Bcachefs Squeezes More Fixes Into Linux 6.12

[royce]

This filesystem has a true opportunity to unseat ZFS at that niche in a way that btrfs has failed to do, glad to see it's got momentum. I don't understand why Torvalds is being such a drama queen about its pace of development.

[geerge]

This filesystem has a true opportunity to unseat ZFS at that niche in a way that btrfs has failed to do, glad to see it's got momentum. I don't understand why Torvalds is being such a drama queen about its pace of development.

Pace of development is fine. Inclusion into mainline in a way that is abnormal to the way things are mainlined, not fine. He's been given an insane amount of leeway, Kent is a lucky boy.

[rmfx]

I don't care about the drama, what I care is that bcachefs has the potential of being the best Linux file system overall.
When it's stabilized and optimized even further, I will give it a try with excitement, and adopt it if it's as good as promised.

[fitzie]

wheres my drama?

kent hasn't started any drama so there hasn't been any. it's a good sign for sure, although much of the linus drama come during the start of the release cycle, so we getting close to it happening if it is going to happen. but signs are pretty good that the upcoming one will be uneventful, for example there's a lot more activity on the bcachefs mailing lists. as for a touching the filesystem as a user, it seems like it's still bleeding edge with a lot more corner cases to shake out, but the biggest concern for me is that he changes the on disk format automatically in a way that can prevent going to prior kernels, and there's really no documentation of this behavior at all. kent has said that once he's ready to remove the experimental label then things he'll change the filesystem to be fixed requiring administrators to enable format changes.

[pWe00Iri3e7Z9lHOX2Qx]

Are there plans to update the equivalent of a 4 disk RAID 10 layout to be able to survive 2 disk failures like you can in a traditional RAID 10 if the failures are one each in different mirrors? Total loss with any 2nd disk failing is a non-starter for me.

[Polaris]

Bug fixes are always welcome, but it’s also a bit disappointing. Bcachefs was promoted as being more stable than Btrfs and having a cleaner codebase even before it entered the kernel. Now, all we’re seeing are bug fixes after bug fixes, and many of them are very serious bugs. A quick look at the issues on bcachefs’s github issues shows dozens of bugs still running in the wild, which is quite concerning for a filesystem that "won't eat your data". I can only hope it gets truly stable in a few years.

[Developer12]

Failing to clear the queue of operations during a shutdown shouldn't cause problems. System crashes aren't know to clear queues.

If a CoW filesystem can't deal with that, it's not worthy of the title.

[User29]

Acutally appears to be good though... as these are fixes that he generated by testing... which is what he was told to do.

testing (just like reading the documentation) is for the weak

[Raka555]

I was thinking of giving it a try on a second backup server.
That should put it through its paces without me risking losing any data. BTW that is a test btrfs could never pass. Always got some corruption after a few months.

But reading about "inodes hanging" made me reconsider. Maybe I will wait for a few more kernels to pass before I try it.

[cakeisamadeupdrug]

If most people cared about something to the extent that Michael cared about Bcachefs the world would be a much more productive place.