1 |
On 28.04.2015 16:53, Stefan G. Weichinger wrote: |
2 |
> |
3 |
> I now booted from a recent sysresccd ... and it didn't assemble/detect |
4 |
> the LVM-PV ... *sigh* |
5 |
> |
6 |
> So I couldn't even mount the LV to chroot in there. |
7 |
> |
8 |
> The mdadm-RAID-devices were there and active. |
9 |
> |
10 |
> I didn't want to screw around with the LVM too much as there is data on |
11 |
> there (yes, there are backups, but ...). |
12 |
> |
13 |
> pvscan always told me something like: |
14 |
> |
15 |
> No device found for PV hMMm0m-w4Ds-SU77-ATjF-iF1A-0qfq-V2IcXb |
16 |
> |
17 |
> .... |
18 |
|
19 |
ok, one step further: |
20 |
|
21 |
(no exact cut and paste because of java-based KVM access) |
22 |
|
23 |
pvscan --cache |
24 |
|
25 |
tells me: |
26 |
|
27 |
Found duplicate PV hMMm0m-w4Ds-SU77-ATjF-iF1A-0qfq-V2IcXb: |
28 |
|
29 |
using /dev/sdj1 not /dev/md3 |
30 |
|
31 |
-- |
32 |
|
33 |
md3 is the correct array to form the PV ... /dev/sdj1 is actually a |
34 |
partition that is an active *member* of the array /dev/md3 |
35 |
|
36 |
I tried to filter this partition by editing lvm.conf without success. |
37 |
|
38 |
Under suse I don't see /dev/sdj1 as pv: |
39 |
|
40 |
# pvdisplay /dev/sdj1 |
41 |
Failed to read physical volume "/dev/sdj1" |
42 |
|
43 |
- |
44 |
|
45 |
Should I |
46 |
|
47 |
* "pvremove /dev/sdj1" from sysresccd |
48 |
|
49 |
* in suse: remove sdj1 from array md3 ... overwrite header and re-add ? |
50 |
|
51 |
Oh my. |
52 |
|
53 |
!! BUT -> |
54 |
|
55 |
I now found another nice fact: |
56 |
|
57 |
/dev/md2 (NO LVM) is unused since the initial installation. |
58 |
|
59 |
I even *documented* that in the project wiki years ago. |
60 |
|
61 |
40 GB of unused space on a RAID1 of SAS hdds. Nice! |
62 |
|
63 |
Sounds like my new rootfs ;-) |
64 |
|
65 |
dracut plus these md-lines should work then. |
66 |
|
67 |
Stefan |