1 |
On Thursday 07 April 2011 06:52:26 BRM wrote: |
2 |
> ----- Original Message ---- |
3 |
> |
4 |
> > From: Joost Roeleveld <joost@××××××××.org> |
5 |
> > |
6 |
> > On Thursday 07 April 2011 06:20:55 BRM wrote: |
7 |
> > > ----- Original Message ---- |
8 |
> > > |
9 |
> > > > From: Neil Bothwick <neil@××××××××××.uk> |
10 |
> > > > |
11 |
> > > > On Thu, 07 Apr 2011 05:22:41 -0500, Dale wrote: |
12 |
> > > > > I want to do it this way because I don't trust LVM enough |
13 |
> > > > > to put my |
14 |
> > > > > |
15 |
> > > > > OS |
16 |
> > > > > on. Just my personal opinion on LVM. |
17 |
> > > > |
18 |
> > > > This doesn't make sense. Your OS can be reinstalled in an hour |
19 |
> > > > or two, your photos etc. are irreplaceable. |
20 |
> > > |
21 |
> > > Makes perfect sense to me as well. |
22 |
> > > |
23 |
> > > Having installed LVM - and then removed it due to issues; namely, |
24 |
> > > the fact that one of the hard drives died taking out the whole LVM |
25 |
> > > group, leaving the OS unbootable, and not easily fixable. There |
26 |
> > > was a thread on that (started by me) a while back (over a year). |
27 |
> > > |
28 |
> > > So, perhaps if I had a RAID to underly so I could mirror drives |
29 |
> > > under LVM |
30 |
> > > |
31 |
> > > for recovery I'd move to it again. But otherwise it is just a PITA |
32 |
> > > waiting |
33 |
> > > |
34 |
> > > to happen. |
35 |
> > > |
36 |
> > > Ben |
37 |
> > |
38 |
> > Unfortunately, any method that spreads a filesystem over multiple disks |
39 |
> > can be |
40 |
> > |
41 |
> > affected if one of those disks dies unless there is some mechanism in |
42 |
> > place that can handle the loss of a disk. |
43 |
> > For that, RAID (with the exception of striping, eg. RAID-0) provides |
44 |
> > that. |
45 |
> > |
46 |
> > Just out of curiousity, as I never had the need to look into this, I |
47 |
> > think that, in theory, it should be possible to recover data from LVs |
48 |
> > that were not |
49 |
> > |
50 |
> > using the failed drive. Is this assumption correct or wrong? |
51 |
> |
52 |
> If you have the LV configuration information, then yes. Since I managed to |
53 |
> find the configuration information, I was able to remove the affected PVs |
54 |
> from the VG, and get it back up. |
55 |
> I might still have it running, but I'll back it out on the next rebuild - or |
56 |
> if I have a drive large enough to do so with in the future. I was wanting |
57 |
> to use LVM as a bit of a software RAID, but never quite got |
58 |
> that far in the configuration before it failed. It does do a good job at |
59 |
> what it's designed for, but I would not trust the OS to it either since the |
60 |
> LVM configuration is very important to keep around. |
61 |
> |
62 |
> If not, good luck as far as I can tell. |
63 |
> |
64 |
> Ben |
65 |
|
66 |
LVM isn't actually RAID. Not in the sense that one gets redundancy. If you |
67 |
consider it to be a flexible partitioning method, that can span multiple disks, |
68 |
then yes. |
69 |
But when spanning multiple disks, it will simply act like JBOD or RAID0. |
70 |
Neither protects someone from a single disk failure. |
71 |
|
72 |
On critical systems, I tend to use: |
73 |
DISK <-> RAID <-> LVM <-> Filesystem |
74 |
|
75 |
The disks are as reliable as Google says they are. They fail or they don't. |
76 |
RAID protects against single disk-failure |
77 |
LVM makes the partitioning flexible |
78 |
Filesystems are picked depending on what I use the partition for |
79 |
|
80 |
-- |
81 |
Joost |