| 1 |
Am Sun, 17 Sep 2017 01:20:45 -0500 |
| 2 |
schrieb Dan Douglas <ormaaj@×××××.com>: |
| 3 |
|
| 4 |
> On 09/16/2017 07:06 AM, Kai Krakow wrote: |
| 5 |
> > Am Fri, 15 Sep 2017 14:28:49 -0400 |
| 6 |
> > schrieb Rich Freeman <rich0@g.o>: |
| 7 |
> > |
| 8 |
> >> On Fri, Sep 8, 2017 at 3:16 PM, Kai Krakow <hurikhan77@×××××.com> |
| 9 |
> >> wrote: |
| 10 |
> [...] |
| 11 |
> >> |
| 12 |
> >> True, but keep in mind that this applies in general in btrfs to any |
| 13 |
> >> kind of modification to a file. If you modify 1MB in the middle |
| 14 |
> >> of a 10GB file on ext4 you end up it taking up 10GB of space. If |
| 15 |
> >> you do the same thing in btrfs you'll probably end up with the |
| 16 |
> >> file taking up 10.001GB. Since btrfs doesn't overwrite files |
| 17 |
> >> in-place it will typically allocate a new extent for the |
| 18 |
> >> additional 1MB, and the original content at that position within |
| 19 |
> >> the file is still on disk in the original extent. It works a bit |
| 20 |
> >> like a log-based filesystem in this regard (which is also |
| 21 |
> >> effectively copy on write). |
| 22 |
> > |
| 23 |
> > Good point, this makes sense. I never thought about that. |
| 24 |
> > |
| 25 |
> > But I guess that btrfs doesn't use 10G sized extents? And I also |
| 26 |
> > guess, this is where autodefrag jumps in. |
| 27 |
> |
| 28 |
> According to btrfs-filesystem(8), defragmentation breaks reflinks, in |
| 29 |
> all but a few old kernel versions where I guess they tried to fix the |
| 30 |
> problem and apparently failed. |
| 31 |
|
| 32 |
It was splitting and splicing all the reflinks which is actually a tree |
| 33 |
walk with more and more extents coming into the equation, and ended up |
| 34 |
doing a lot of small IO and needing a lot of memory. I think you really |
| 35 |
cannot fix this when working with extents. |
| 36 |
|
| 37 |
|
| 38 |
> This really makes much of what btrfs |
| 39 |
> does altogether pointless if you ever defragment manually or have |
| 40 |
> autodefrag enabled. Deduplication is broken for the same reason. |
| 41 |
|
| 42 |
It's much easier to fix this for deduplication: Just write your common |
| 43 |
denominator of an extent to a tmp file, then walk all the reflinks and |
| 44 |
share them with parts of this extent. |
| 45 |
|
| 46 |
If you carefully select what to defragment, there should be no problem. |
| 47 |
A defrag tool could simply skip all the shared extents. A few fragments |
| 48 |
do not hurt performance at all, but what's important is spatial |
| 49 |
locality. A lot small fragments may hurt performance a lot, so one |
| 50 |
could give the defragger a hint when to ignore the rule and still |
| 51 |
defragment the extent. Also, when your deduplication window is 1M you |
| 52 |
could probably safely defrag all extents smaller than 1M. |
| 53 |
|
| 54 |
|
| 55 |
-- |
| 56 |
Regards, |
| 57 |
Kai |
| 58 |
|
| 59 |
Replies to list-only preferred. |
Archives