Btrfs Brings "Pretty Beefy" Changes In Linux 3.2

[phoronix]

Phoronix: Btrfs Brings "Pretty Beefy" Changes In Linux 3.2

The pull request for the Btrfs file-system in the Linux 3.2 kernel has finally come in this Sunday. It brings some fairly significant changes for this up-and-coming Linux file-system...

http://www.phoronix.com/vr.php?view=MTAxMjM

[macemoneta]

btrfsck

Without a working btrfsck, all the patches in the world won't convince me to put any data into a btrfs filesystem. I've been familiarizing myself with btrfs in a virtual system, and it's great. I just wish the userspace tools were in a usable state.

[Hamish Wilson]

Without a working btrfsck, all the patches in the world won't convince me to put any data into a btrfs filesystem. I've been familiarizing myself with btrfs in a virtual system, and it's great. I just wish the userspace tools were in a usable state.

Seconded. I would not go near a file-system without fsck capabilities with a thirty-foot pole. No matter how much more advanced it is. But I wish them luck in getting there.

[GreatEmerald]

I'm definitely looking forward to 3.2. I'm already using Btrfs (haven't had any situation where I would need a check tool so far, and the only feature I miss so far is the support for disk labels), and the improvements in psb-gfx are very much useful.

Hmm, the new numbering system is rather confusing for me right now, since I'm used to two numbers before the stable release number, but I guess it makes sense, as the second number wouldn't add anything useful.

[oxplot]

Without a working btrfsck, all the patches in the world won't convince me to put any data into a btrfs filesystem. I've been familiarizing myself with btrfs in a virtual system, and it's great. I just wish the userspace tools were in a usable state.

From what I know about COW file systems (e.g. ZFS), if the intermediate layers between the FS and the physical media handle flushes/barriers correctly (unlike dm-crypt module), then no crash should leave the FS in an inconsistent state. This might be one of the reasons why btrfsck utility is not yet released. This is not to say btrfsck is not needed at all, but for most common scenarios, the COW nature of the FS should make it resilient to corruptions. I have used btrfs for the past 14 months and had numerous crashes due to system freezes and power cuts. Except one time that I had the btrfs mounted on an encrypted block device, the FS mounted with no issues every time. However in that one time that it did crash, I lost everything! I wish there was btrfsck so that I could save *some* of my files.

[gilboa]

Seconded. I would not go near a file-system without fsck capabilities with a thirty-foot pole. No matter how much more advanced it is. But I wish them luck in getting there.

Fully agree.
Given the fact that making btrfs the default in F17 [1] depends [2] on having a stable fsck utility, the lack of progress on this end may push it to F18 or even F19.

- Gilboa
[1] https://bugzilla.redhat.com/show_bug.cgi?id=689509
[2] https://bugzilla.redhat.com/show_bug.cgi?id=689512

[RavFX]

I stopped using BTRFS last Friday. The accumulation of silent corruption made me mad and I switched everything to Reiserfs (I will switch everything to reiser4 once I will have completed my patch for 3.1)

I started to migrate other computers I support out of btrfs too... no fsck = bye bye btrfs.

[Artemis3]

Why the abandoned reiserfs and not ext4 which is currently leading benchmarks and will allow future migration into btrfs?

I'm currently using ext4 simply because btrfs is not ready, even with an fsck tool it needs a bit of time and lots of usage to uncover hidden nasties.

How could you not expect to lose all your data using an unfinished fs? You are supposed to test it, not use it for production...

[RavFX]

Why the abandoned reiserfs and not ext4 which is currently leading benchmarks and will allow future migration into btrfs?

I'm currently using ext4 simply because btrfs is not ready, even with an fsck tool it needs a bit of time and lots of usage to uncover hidden nasties.

How could you not expect to lose all your data using an unfinished fs? You are supposed to test it, not use it for production...

I don't use EXT file system except ext2 for boot. I saw just too many destroyed FS for stupid reason with those. Benchmark is not my priority, my data are more important than that.

As using unfinished fs, yeah, I was testing it and it was not in production but... A FSCK tools is not too much to ask and would have probably fixed my problems.. they promised one in "7 weeks" one years ago but we still don't see it so during that time the silent corruption stack-up slowly and silently... (I don't crash the computer, it just do that himself).

And now they said thing as they will keep the FSCK for Oracle Linux only.

I have backup on external USB hard drive, But I like to not have to use them...

[sbergman27]

...I switched everything to Reiserfs (I will switch everything to reiser4 once I...

Note that Hans has moved again. Send your bug reports to:

Pleasant Valley State Prison
24863 West Jayne Avenue
Coalinga, CA 93210

Update: Oops. All mail sent to the physical address of the prison will be returned to sender. Send bug reports to:

P.O. Box 8500
Coalinga, CA 93210