1 |
On Fri, 2008-02-15 at 21:05 +0100, Wael Nasreddine wrote: |
2 |
> Currently I have 2 partitions, a root and home partition, fortunately |
3 |
> on LVM array, I was thinking of splitting them to "/, /usr, /var, /home, |
4 |
> /usr/portage, /mnt/storage" the latter is to be used for Mp3z (around |
5 |
> 12000) and movies... |
6 |
> |
7 |
> I was thinking of having the below filesystem schema: |
8 |
> / : ext3 (-j -O dir_index,sparse_super,filetype) (Good mkfs options ??) |
9 |
> /usr : xfs (I never used it so please suggest mkfs.xfs options) |
10 |
> /var : // |
11 |
> /home : ext3 (-m 0 -j -O dir_index,sparse_super,filetype) (Good mkfs options ??) |
12 |
> /usr/portage : ReiserFS (3? 4? options??) |
13 |
> /mnt/storage : ext3 (-m 0 -j -O dir_index,sparse_super,filetype) (Good mkfs options ??) |
14 |
> |
15 |
> |
16 |
> Could you please comment/complete/change the schema above ?? I really |
17 |
> would like to speed up my system a little bit, My system is entirely |
18 |
> built on LVM array, and LVM is on DM-CRYPT so as you can see it's a |
19 |
> quite slow due to the encryption... |
20 |
> |
21 |
> Oh one last thing, What do you suggest for a server? I have a Gentoo |
22 |
> server and uptime can be over 5/6 months, everytime I reboot the |
23 |
> server I have to manually scan the filesystem due to errors |
24 |
> everywhere, any suggestions?? |
25 |
> |
26 |
> Thanks... |
27 |
|
28 |
First of all, if there are filesystem errors, check your cables, your |
29 |
controller and your disks. I don't think filesystem errors count as |
30 |
normal behavior ... |
31 |
|
32 |
To your filesystem scheme: Why do you use xfs for usr? AFAIK XFS is good |
33 |
at write speed but not worth the trouble when reading data and data in |
34 |
usr is usually written once, updated every few months and read many |
35 |
times a week (on rebooting Desktop PCs maybe once a day). I'd use |
36 |
reiserfs3.6, maybe even without notail to make it more space efficient. |
37 |
|
38 |
I'd also use ext2 on /usr/portage. These data don't need journaling. |
39 |
Everything's got an MD5-sum to make sure it's unchanged after a crash |
40 |
and you can easily resync. I found ext2 with 2k blocks to be faster than |
41 |
reiserfs3.6, even on read-performance. |
42 |
|
43 |
If I were you, I'd also use separate volumes for /tmp and /var/tmp |
44 |
(without ccache) with xfs. |
45 |
|
46 |
/home could use data=journal. Those data are precious and if I remember |
47 |
correctly, this setting even brings an obscure (i.e. undocumented) speed |
48 |
improvement with many parallel disk accesses, for example in a |
49 |
multi-user environment. |