An update on bcachefs [LWN.net]

An update on bcachefs

Posted May 24, 2018 1:21 UTC (Thu) by simcop2387 (subscriber, #101710)
In reply to: An update on bcachefs by rahvin
Parent article: An update on bcachefs

I think a lot of the bad rap came not from the btrfs developers themselves, but a few distros that started supporting it in production a bit too early. This caused a number of people to start using it and then experience those corner cases. It got a lot of testing but left a bad taste in people's mouths.

to post comments

An update on bcachefs

Posted May 24, 2018 10:30 UTC (Thu) by pizza (subscriber, #46) [Link] (1 responses)

"Corner cases" my right foot. The reason for the bad taste is that btrfs would completely eat itself even when using supposedly stable/fully-supported configurations on stable hardware.

I've had two [might-as-well-be-]total filesystem losses with btrfs, both after clean shutdown/reboot cycles on lightly-loaded, battery-backed hardware that hasn't so much as hiccupped before or since.

An update on bcachefs

Posted May 25, 2018 15:54 UTC (Fri) by drag (guest, #31333) [Link]

There are two major times my personal Linux boxes ate my data...

1st time was when my cat yanked out a PCI card that wasn't screwed in. Corrupted xfs metadata, had to sort through tens of thousands of files in the recovery bin.

2nd time was while using btrfs.

There was also a couple times while using btrfs on external storage device I expanded it wedged itself into a corner. Ran out of inodes or something like that. It was a long time ago. Didn't lose data, but the file system was effectively rendered worthless and needed to be rebuilt.

Now these things are all partially my fault, but I have used Linux on garbage-level (sometimes literally) hardware for years and it's been proven to be tough.

I am looking forward to bcachefs getting support from various distro installers.

An update on bcachefs

Posted May 24, 2018 13:11 UTC (Thu) by dcg (subscriber, #9198) [Link]

I don't think being unstable was really that much of a problem at first; people expected btrfs to be unstable anyway. The problem is that it kept being unstable because there weren't developers that would focus on fixing that. There were developers that worked on adding features (often not cleanly, eg qgroups were very unstable and still today they impact performance so much that people recommends not using them unless it's necessary), but the core was left unattended. Which is the reason why, still today and after so many years, btrfs does not have clean, ZFS-class RAID5/6 support, or integrated encryption. The lack of improvement in some key fields seem to be making people feel that btrfs is stagnating, and who would blame them? It is in fact one of the motivations behind bcachefs.

As someone who has been using btrfs for many years without problems I also feel that btrfs is not appreciated enough (and that bcachefs is over-hyped - Overstreet has criticized btrfs for its stability, but I have seen several cases of people having corruption issues in the bcachefs IRC channel). But I wonder how much can that last if btrfs keeps not progressing in basic features like RAID5/6?