Bcachefs Repair Code Reaching Complete & Robust Recovery

[anarki2]

It's still marked experimental for a reason. You aren't supposed to be using it for anything you find super important. Distro installers don't offer it. You need to go out of your way to set up a storage pool with it. There are tons of issues that were never going to be found without enough people beating on it every day, and that wasn't going to happen until it was in mainline.

This. The whole reason for having an "experimental" new feature is to have people experiment with it. You use it at your own risk. It's not like suddenly it becomes default in existing distros with kernels that don't even have it, then even automatically converts your existing FSes. Some people here treat Linux kernel like a baby, but they just don't understand how development works. How do you make something stable if no one is using it?

For the record, here's an explainer on what a module flagged EXPERIMENTAL is:

EXPERIMENTAL — Prompt for development and/or incomplete code/drivers

Some of the many things that Linux supports (such as network drivers, file systems, network protocols, etc.) can be in a state of development where the functionality, stability, or the level of testing is not yet high enough for general use. This is usually known as the "alpha-test" phase among developers. If a feature is currently in alpha-test, the developers usually discourage uninformed widespread use of this feature by the general public to avoid "Why doesn't this work?" mail messages. However, active testing and use of these systems is welcomed. Just be aware that it may not meet the normal level of reliability or may fail to work in some special cases. Detailed bug reports from people familiar. with the kernel internals are usually welcomed by the developers. (

[ehansin]

I have been diving increasingly deeper into filesystem and storage stuff. I myself am happy to see Bcachefs make mainline. Obviously we want good developer communication happening, but I don't care to focus too much on the drama without knowing everything. I do think finally getting into mainline, but also marked as EXPERIMENTAL allows more people to play with Bcachefs and pound away at it, see what does and what does not work. I've mentioned before, but I watched Kent's ~1 hour long video he did explaining the architecture and how things were structured. Admittedly a lot over my head, but I came away pretty impressed. Time will tell with Bcachefs, but happy to see it around.

[lsatenstein]

I am awaiting gparted's inclusion of bcachefs. before I switch over. If zfs was partition based, in lieu of device based, I would have standardized on it.
I am a Desktop user, and I write code. My stuff fits into 60gigs of disk space. zfs was overkill.

[skeevy420]

I am awaiting gparted's inclusion of bcachefs. before I switch over. If zfs was partition based, in lieu of device based, I would have standardized on it.
I am a Desktop user, and I write code. My stuff fits into 60gigs of disk space. zfs was overkill.

You can use ZFS with partitions. It's not the ideal configuration, but it's entirely possible. My first ZFS use was a single disk split with NTFS and ZFS. Heck, on Linux you have to either use a partition or have a disk just for a bootloader to have a ZFS root. I use two partitions with my ZFS root.

I use ZFS precisely because it doesn't use partitions in regards to volume management. I don't have to worry about preallocating drive space with things like /usr and /var because they need different compression settings and permissions. To me, thin provisioning and partitions it isn't an ideal setup to work with.

[skeevy420]

I have been diving increasingly deeper into filesystem and storage stuff. I myself am happy to see Bcachefs make mainline. Obviously we want good developer communication happening, but I don't care to focus too much on the drama without knowing everything. I do think finally getting into mainline, but also marked as EXPERIMENTAL allows more people to play with Bcachefs and pound away at it, see what does and what does not work. I've mentioned before, but I watched Kent's ~1 hour long video he did explaining the architecture and how things were structured. Admittedly a lot over my head, but I came away pretty impressed. Time will tell with Bcachefs, but happy to see it around.

The deeper I dove the more I found myself wanting to use two specific file systems: ZFS and HAMMER2. One of those is limited to one specific BSD while the other one can be used on at least 5 operating systems that I know of: Linux, BSD, macOS, Windows, and Solaris. I picked the one that can be used with 5 operating systems.

[varikonniemi]

since this brings no new use time features, only repair code to handle previous bugs, i tihnk it might be acceptable to have such new code appear such late in merge window, especially and only because the filesystem is still EXPERIMENTAL flagged.

[intelfx]

since this brings no new use time features, only repair code to handle previous bugs, i tihnk it might be acceptable to have such new code appear such late in merge window, especially and only because the filesystem is still EXPERIMENTAL flagged.

It's stil new code.

The point of the release cadence and the -rc releases is not to block users from getting features, but to allow time for the _code_ to stabilize. So the only thing that matters is that it's new (untested) code, not whether it's a new user-facing feature.

[varikonniemi]

It's stil new code.

The point of the release cadence and the -rc releases is not to block users from getting features, but to allow time for the _code_ to stabilize. So the only thing that matters is that it's new (untested) code, not whether it's a new user-facing feature.

Things should work differently in normal and experimental code. In experimental, the priority is to just get it out to testers. So the only thing to enforce there is to have it build cleanly, IMO

[intelfx]

Things should work differently in normal and experimental code. In experimental, the priority is to just get it out to testers. So the only thing to enforce there is to have it build cleanly, IMO

Yes, perhaps you have a point wrt. "experimental" features, and on this point I'd be inclined to agree with you. But I'm not sure whether Linux development process (or Linus himself) recognizes this exception.

[ehansin]

The deeper I dove the more I found myself wanting to use two specific file systems: ZFS and HAMMER2. One of those is limited to one specific BSD while the other one can be used on at least 5 operating systems that I know of: Linux, BSD, macOS, Windows, and Solaris. I picked the one that can be used with 5 operating systems.

I am using ZFS on FreeBSD (technically TrueOS) at work. Also have played around with it in virtual environments. Not a power-user, but am a fan. That said, I am happy to give Bcachefs a chance to prove itself. I do get that between licensing and Linux-ism dependencies (I am guessing Bcache is a Linux-only thing, but don't know nor know exactly how Bcachefs it is licensed), getting to work in/on non-Linux operating systems may not be possible. But a solid "in-tree" Linux option (I cannot comment on Btrfs, so don't want to go there - but have used in some cases) would be great.