| 1 |
On Fri, Feb 24, 2012 at 6:31 PM, Richard Yao <ryao@×××××××××××××.edu> wrote: |
| 2 |
>> Am I the only paranoid person who moves them rather than unlinking |
| 3 |
>> them? Oh, if only btrfs were stable... |
| 4 |
> |
| 5 |
> Is this a reference to snapshots? You can use ZFS for those. The |
| 6 |
> kernel modules are only available in the form of 9999 ebuilds right |
| 7 |
> now, but they your data should be safe unless you go out of your way |
| 8 |
> to break things (e.g. putting the ZIL/SLOG on a tmpfs). Alternatively, |
| 9 |
> there is XFS, which I believe also supports snapshots. |
| 10 |
> |
| 11 |
|
| 12 |
FWIW, I'll second the ZFS > btrfs suggestion. I understand people want |
| 13 |
to go btrfs cause its the Linux way but in real world usage, its |
| 14 |
performance is abysmal We've had over a dozen developers switch to |
| 15 |
btrfs in my group on their various environments (OpenSUSE, Fedora, own |
| 16 |
rolled distros) and they've all gone back to their previous filesystem |
| 17 |
of choice. |
| 18 |
|
| 19 |
Simplest test I can suggest to btrfs users to attempt is the following: |
| 20 |
|
| 21 |
dd if=/dev/zero of=/mnt/btrfs/file bs=4k count=100 oflag=direct |
| 22 |
dd if=/dev/zero of=/mnt/ext4/file bs=4k count=100 oflag=direct |
| 23 |
|
| 24 |
It will emulate the similar operation to an fdatasync(). |
| 25 |
|
| 26 |
-- |
| 27 |
Doug Goldstein |
Archives