Bcachefs Lands Another Round Of Fixes For Linux 6.7

[phoronix]

Phoronix: Bcachefs Lands Another Round Of Fixes For Linux 6.7

On Friday another round of fixes were merged for the Bcachefs file-system for the in-development Linux 6.7 kernel...

https://www.phoronix.com/news/Bcachefs-More-Fixes-Linux-6.7

[LinAdmin]

>"... and the developers being aware of areas for (performance) improvements moving ahead"

IMHO highest priority should be robustness and the possibility of Scrubbing with automatic repair.

[pipe13]

>"... and the developers being aware of areas for (performance) improvements moving ahead"

IMHO highest priority should be robustness and the possibility of Scrubbing with automatic repair.

...username checks out., although tbf the crc32 hardware enablement shouldn't be real hard once they figure it out.

[HD7950]

At this point the priority should not be performance. To build a good reputation, the main thing is to focus on reliability, something that Btrfs never achieved after more than a decade. A high score in a performance test is of no use if it is not backed up by bombproof endurance. The most important thing about a file system, for the average user, is to forget that it exists.

It doesn't matter if it is at turtle speed but please: DO NOT CORROMPT OR LOSE MY DATA.

[AlanTuring69]

They seem to be dealing both with reliability and performance improvements since it's very much a work in progress. Making the current bcachefs iron clad right now is fairly pointless since the optimizations and performance improvements will change the state of the code, probably significantly. Make no mistake though they are focusing on bugs too. If you use it on a production system that's on you, and the sooner it is relevant in terms of performance the more people will use and report more bugs which then get fixed which gets more users etc. There are plenty of non-critical use-cases when you can throw in a not-necessarily-perfectly-stable filesystem if it has some feature that others don't have.

Nice to see great hygiene on this merge with no drama and hopefully this continues.

[User29]

At this point the priority should not be performance. To build a good reputation, the main thing is to focus on reliability

This is true. However... Building reputation takes a lot of time and is booooooriiiiiiiing. Winning different benchmarks is probably a bit easier.

[oleid]

A high score in a performance test is of no use if it is not backed up by bombproof endurance. The most important thing about a file system, for the average user, is to forget that it exists.

It doesn't matter if it is at turtle speed but please: DO NOT CORROMPT OR LOSE MY DATA.

It may not corrupt the data, that for sure. But if it is turtle speed few people will use it. And it might turn out that the disk format needs a change to reach faster speeds.

In an ideal world the filesystem should not be too far off from the rest of the filesystems and provide reliability.

[Joe2021]

At this point the priority should not be performance. To build a good reputation, the main thing is to focus on reliability...

It doesn't matter if it is at turtle speed but please: DO NOT CORROMPT OR LOSE MY DATA.

So true, but what was the first thing that happened at this very site when Bcachefs became part of the linux kernel? A benchmark comparing the performance.

Don't get me wrong, I fully understand that this is what phoronix is primary about, but it still demonstrates that most people glance at the performance numbers.

[Vermilion]

the main thing is to focus on reliability, something that Btrfs never achieved after more than a decade

If only Fedora, Meta, SUSE, Google, Synology all took advices from you

[fitzie]

At this point the priority should not be performance. To build a good reputation, the main thing is to focus on reliability, something that Btrfs never achieved after more than a decade. A high score in a performance test is of no use if it is not backed up by bombproof endurance. The most important thing about a file system, for the average user, is to forget that it exists.

It doesn't matter if it is at turtle speed but please: DO NOT CORROMPT OR LOSE MY DATA.

I think their priority needs to be the features. realistically, they will never get reliability until a hyperscaler starts a big test deployment of bcachefs to start to put bcachefs through it's paces, which will expose all the easy to trigger bugs inside the filesystem. and unless they have a minimum set of features, there's little chance of that.

I'll certainly be putting bcachefs through some tests, as I'd like to see it be fulfill it's promise, but I won't bother putting it in a major role until there's a track record, and the quickest way to accelerate that track record is with hyperscalers.