| 1 |
On 10/03/2011 11:47 PM, Pandu Poluan wrote: |
| 2 |
> Hello people! |
| 3 |
> |
| 4 |
> Now, I have the same question as this guy: |
| 5 |
> |
| 6 |
> https://bbs.archlinux.org/viewtopic.php?id=66651 |
| 7 |
> |
| 8 |
> I.e., what is the most robust filesystem for Linux? |
| 9 |
> |
| 10 |
> The box will be used as a gateway/firewall for a branch office, so I |
| 11 |
> really couldn't care less about filesystem performance. But the utility |
| 12 |
> power there is horrendous, so I need something that can shrug off a |
| 13 |
> catastrophic power loss, and/or very fast fsck. |
| 14 |
> |
| 15 |
> I'd also appreciate any tips on mount options to further enhance robustness. |
| 16 |
|
| 17 |
Journaling filesystems are as safe as you'll get on commodity hardware. |
| 18 |
I would choose ext4 because I'm familiar with it, but I'm sure others |
| 19 |
offer the same options. It's also got the fastest fsck that I'm aware of. |
| 20 |
|
| 21 |
From `man tune2fs`: |
| 22 |
|
| 23 |
journal_data |
| 24 |
When the filesystem is mounted with journalling |
| 25 |
enabled, all data (not just metadata) is committed into |
| 26 |
the journal prior to being written into the main |
| 27 |
filesystem. |
| 28 |
|
| 29 |
|
| 30 |
block_validity |
| 31 |
|
| 32 |
(I haven't used this, but spotted it in the man page) |
| 33 |
|
| 34 |
The file system will be mounted with the block_validity |
| 35 |
option enabled, which causes extra checks to be per‐ |
| 36 |
formed after reading or writing from the file system. |
| 37 |
This prevents corrupted metadata blocks from causing |
| 38 |
file system damage by overwriting parts of the inode |
| 39 |
table or block group descriptors. This comes at the |
| 40 |
cost of increased memory and CPU overhead, so it is |
| 41 |
enabled only for debugging purposes. (This option is |
| 42 |
currently only supported by the ext4 file system driver |
| 43 |
in 2.6.35+ kernels.) |
Archives