| 1 |
On Wed, 2 Jun 2004 09:57:13 +0300 |
| 2 |
Dadi <thewalrus@××××××××××××××.org> wrote: |
| 3 |
|
| 4 |
> Regarding reiser, some friends of mine were telling me that it's |
| 5 |
> notorious for breaking files when unexpected things happen. Is this |
| 6 |
> true? I mean I had my share of trouble with reiser also on other |
| 7 |
> harddisks but always lived with the thought that it was a hardware |
| 8 |
> problem more then a fs problem. Now it's all coming together and |
| 9 |
> starting to make sense. |
| 10 |
> |
| 11 |
|
| 12 |
There seems to be a common misconception regarding journaling filesystems. |
| 13 |
Normally, they only guarantee filesystem inegrity, *not* data integrity as |
| 14 |
many seem to believe. |
| 15 |
So, if your machine crashes, unflushed data is lost and the affected |
| 16 |
files are corrupted. The filesystem however will be correct after log |
| 17 |
replay, so all files which were closed or flushed to disk at crash time |
| 18 |
are unaffected. |
| 19 |
This is true for reiserfs, jfs and xfs. ext3 is slightly better when using |
| 20 |
the data=ordered option and it can even guarantee data integrity through |
| 21 |
the data=journal option. However this severly decreases write performance. |
| 22 |
|
| 23 |
It is important to note, that all journaling filesystems rely on data |
| 24 |
ordering. If you are using your hard-disk's write cache, things can break |
| 25 |
badly. This is probably the main reason for reiserfs' bad reputation. |
| 26 |
|
| 27 |
Regards |
| 28 |
|
| 29 |
-- |
| 30 |
gentoo-security@g.o mailing list |
Archives