Any experiences with BTRFS on Fedora 17/Fedora 18?

Rupadhya · 01-29-2013, 02:51 PM

I was looking at the documentation for BTRFS and wondering if there is any reason to migrate to it as of yet. I think it is still experimental in the kernel, and most of my data is JPGs of various sizes.

My hard drive topology include:

300 Gig formatted as EXT4
1 Terabyte formatted as EXT4, mostly for backup of the 300 Gig.

It looks like BTRFS is really good for multi-volume RAID configurations, and really big file systems. I might be wrong in my assumptions.

- Raj

Emerson · 01-29-2013, 04:29 PM

I'd use XFS for this purpose for now.

Rupadhya · 01-29-2013, 04:37 PM

From reading wikipedia.org, it looks more baked in then BTRFS. Thanks for the feedback.

- Raj

Rupadhya · 04-26-2013, 06:56 PM

I wanted to fsck the drive to check if I corrupted it, and ext4 fsck exists on my machine, but I don't have a fsck.btrfs. Does it exist?

- Raj

Code:

# fsck /home/julie/btrfs
fsck from util-linux 2.22.2
fsck: fsck.btrfs: not found
fsck: error 2 while executing fsck.btrfs for /dev/sdf3
# fsck /home/julie/ext4
fsck from util-linux 2.22.2
e2fsck 1.42.5 (29-Jul-2012)
EXT4: clean, 9640/1925120 files, 7031100/7679743 blocks

guyonearth · 04-27-2013, 07:23 PM

Quote:

Originally Posted by Rupadhya

I wanted to fsck the drive to check if I corrupted it, and ext4 fsck exists on my machine, but I don't have a fsck.btrfs. Does it exist?

- Raj

Code:

# fsck /home/julie/btrfs
fsck from util-linux 2.22.2
fsck: fsck.btrfs: not found
fsck: error 2 while executing fsck.btrfs for /dev/sdf3
# fsck /home/julie/ext4
fsck from util-linux 2.22.2
e2fsck 1.42.5 (29-Jul-2012)
EXT4: clean, 9640/1925120 files, 7031100/7679743 blocks

You use btrfsck, but it's still experimental.

Rupadhya · 04-27-2013, 08:03 PM

Thank you for your response.

Something is not right on my machine, though.

Code:

# mount | grep btrfs
/dev/sdf3 on /home/julie/btrfs type btrfs (rw,relatime,seclabel,space_cache)
# df -h
Filesystem      Size  Used Avail Use% Mounted on
devtmpfs        996M     0  996M   0% /dev
tmpfs          1005M  6.5M  998M   1% /dev/shm
tmpfs          1005M  8.0M  997M   1% /run
tmpfs          1005M     0 1005M   0% /sys/fs/cgroup
/dev/sda3        50G   11G   36G  23% /
tmpfs          1005M  4.6M 1000M   1% /tmp
/dev/sda1       477M   88M  365M  20% /boot
/dev/sda5       222G  178G   33G  85% /home
//rajhome1/h    694G  435G  224G  66% /home/raj/c
/dev/sdf1        24G  1.1G   23G   5% /home/julie/Pictures
/dev/sdf2        29G   27G  1.1G  97% /home/julie/ext4
/dev/sdf3        22G  1.1G   19G   6% /home/julie/btrfs
# btrfsck /home/julie/btrfs
No valid Btrfs found on /home/julie/btrfs
# uname -r
3.9.0-rc8

Any ideas?

- Raj

salasi · 04-28-2013, 06:45 AM

You might want to have a look at this article. This is an interesting article by an Australian ISP who have made an experiment with BTRFS, just to see how things are coming along.

Quote:

It looks like BTRFS is really good for multi-volume RAID configurations, and really big file systems.

Well, yes there are certainly some advantages to this kind of approach (the combined 'LVM + filesystem' part, which you could also describe as 'ZFS-like', if you wanted) for 'big data'. On the other hand, 'big data' applications tend to be more unforgiving of data snafus, and so newly developed filesystem may well have difficulty getting adopted in this area, due to the difficulty achieving the required 'tried 'n tested' status without adventurous early adopters driving the project forward (Would you, eg, take much confidence from the idea that your hosting provider had decided to keep your data safe by keeping all on a freshly minted filesystem? Many people wouldn't.)

The 'snapshotting' part seems potentially useful to many users, whether they be desktop users or 'big data' users, even if the non-geeky probably won't see the advantage until it is made directly apparent to them.

Potentially, BTRFS could have advantages in different areas, from big to small, but the advantages that different users want from it are probably different, so the testing issue, and demonstrating the practical reliability, will probably take a bit of time.