1 |
On Saturday 21 March 2009 20:39:08 Jarry wrote: |
2 |
> Alan McKinnon wrote: |
3 |
> >> cp -a /mnt/gentoo/backup/var /mnt/gentoo/var |
4 |
> >> cp -a /mnt/gentoo/backup/usr /mnt/gentoo/usr |
5 |
> > |
6 |
> > Um, no. This gives you new usr and var directories like so: |
7 |
> > /usr/usr/ |
8 |
> > /var/var |
9 |
> > |
10 |
> > You want: |
11 |
> > cp -a /mnt/gentoo/backup/var /mnt/gentoo/ |
12 |
> > cp -a /mnt/gentoo/backup/usr /mnt/gentoo/ |
13 |
> |
14 |
> Thanks for correction! |
15 |
> |
16 |
> > With lvm, this becomes a breeze. |
17 |
> |
18 |
> I remember having lvm2 a few years ago, and despite of that I could not |
19 |
> extend any partition, which was being used. What is then lvm2 good for, |
20 |
> if I can not extend partitions on-the-fly? I can not unmount /usr before |
21 |
> extending... |
22 |
|
23 |
That is not lvm's fault, it is the fault of the OS. |
24 |
|
25 |
/usr is not a filesystem that changes much anyway. If you look at a few |
26 |
similar machines, you can guess quite accurately what it's size is going to |
27 |
be. |
28 |
|
29 |
/var, database directories, home directories - these are the things you can |
30 |
change on the fly. These are also the things that you do want to change on the |
31 |
fly. |
32 |
|
33 |
> And one more counter-argument: with traditional partitions I can select |
34 |
> where a certain partition is (physically). Those partitions accessed |
35 |
> frequently I put to the beginning of the disk with higher transfer-rate. |
36 |
> In my case, it makes quite difference: |
37 |
> |
38 |
> obelix ~ # hdparm -t /dev/md2 |
39 |
> Timing buffered disk reads: 252 MB in 3.02 seconds = 83.23 MB/sec |
40 |
> |
41 |
> obelix ~ # hdparm -t /dev/md9 |
42 |
> Timing buffered disk reads: 150 MB in 3.02 seconds = 49.72 MB/sec |
43 |
|
44 |
You have no guarantee whatsoever that the data resides on the part of the disk |
45 |
you think it resides on, so this entire argument becomes moot. Today, by happy |
46 |
coincidence, it is. Tomorrow with another drive it might not be. You also have |
47 |
to deal with the effect of disk caching. And you didn't do the only real test |
48 |
the remotely means anything at all - random writes. Throughout measurements |
49 |
are meaningless as the thing you measure hardly ever happens in real life. |
50 |
It's a lot like determining the suitability of a future wife by measuring her |
51 |
foot size: a perfectly correct measure, and also a perfectly useless one. |
52 |
|
53 |
It's this kind of thinking that keeps people trapped in circumstance and |
54 |
unable to take advantage of new ideas. In the IT industry, it is rife. |
55 |
|
56 |
-- |
57 |
alan dot mckinnon at gmail dot com |