| 1 |
Am 13.08.2012 16:52, schrieb Michael Mol: |
| 2 |
> On Mon, Aug 13, 2012 at 10:42 AM, Michael Hampicke |
| 3 |
> <mgehampicke@×××××.com <mailto:mgehampicke@×××××.com>> wrote: |
| 4 |
> |
| 5 |
> Have you indexed your ext4 partition? |
| 6 |
> |
| 7 |
> # tune2fs -O dir_index /dev/your_partition |
| 8 |
> # e2fsck -D /dev/your_partition |
| 9 |
> |
| 10 |
> Hi, the dir_index is active. I guess that's why delete operations |
| 11 |
> take as long as they take (index has to be updated every time) |
| 12 |
> |
| 13 |
> |
| 14 |
> 1) Scan for files to remove |
| 15 |
> 2) disable index |
| 16 |
> 3) Remove files |
| 17 |
> 4) enable index |
| 18 |
> |
| 19 |
> ? |
| 20 |
> |
| 21 |
> -- |
| 22 |
> :wq |
| 23 |
|
| 24 |
Other things to think about: |
| 25 |
|
| 26 |
1. Play around with data=journal/writeback/ordered. IIRC, data=journal |
| 27 |
actually used to improve performance depending on the workload as it |
| 28 |
delays random IO in favor of sequential IO (when updating the journal). |
| 29 |
|
| 30 |
2. Increase the journal size. |
| 31 |
|
| 32 |
3. Take a look at `man 1 chattr`. Especially the 'T' attribute. Of |
| 33 |
course this only helps after re-allocating everything. |
| 34 |
|
| 35 |
4. Try parallelizing. Ext4 requires relatively few locks nowadays (since |
| 36 |
2.6.39 IIRC). For example: |
| 37 |
find $TOP_DIR -mindepth 1 -maxdepth 1 -print0 | \ |
| 38 |
xargs -0 -n 1 -r -P 4 -I '{}' find '{}' -type f |
| 39 |
|
| 40 |
5. Use a separate device for the journal. |
| 41 |
|
| 42 |
6. Temporarily deactivate the journal with tune2fs similar to MM's idea. |
| 43 |
|
| 44 |
Regards, |
| 45 |
Florian Philipp |
Archives