Hi Adam,
Thanks a lot for your solution, I understand the way.
I will try it the next week, I'll wait for the return of Maor on 12th, to know his
advice.
Kind regards
Alain
Alain VONDRA
Chargé d'exploitation des Systèmes d'Information
Direction Administrative et Financière
+33 1 44 39 77 76
UNICEF France
3 rue Duguay Trouin 75006 PARIS
www.unicef.fr
-----Message d'origine-----
De : Adam Litke [mailto:alitke@redhat.com]
Envoyé : vendredi 10 avril 2015 17:56
À : VONDRA Alain
Cc : users(a)ovirt.org
Objet : Re: [ovirt-users] Storage domain not in pool issue
On 10/04/15 08:46 +0000, VONDRA Alain wrote:
Hi Adam,
Ok, if it's a safe solution, I can try it.
Can you give me the commands to do the job, I am not so familiar with LVM.
Here is a start. This will help you scrape off your VM disks into files. I don't
have time to write up the process of importing them back into oVirt but you'll likely
have to use virt-manager to create a libvirt VM using the disks you've extracted and
then use virt-v2v to import it into an oVirt storage domain. This process will leave your
storage domain in-tact so it's safe and if someone comes up with an easier solution
you can try that later.
0. Place one vdsm host into maintenance mode
1. (ENGINE HOST) Determine which imageIDs correspond to which VMs in your storage domain
psql -U <dbUser> <dbName> -c "select storage_id, image_group_id, vm_names
from all_disks where storage_id = 'd7b9d7cc-f7d6-43c7-ae13-e720951657c9';"
2. (VDSM HOST) Connect to the storage
sudo vdsClient -s 0 connectStorageServer 3 <old SPUUID> \
id=`uuidgen`,connection=<iSCSI_IP>,iqn=<iqn>,tpgt=<tpgt>,port=3260,user=,password=
3. (VDSM HOST) Wait for connection to complete by checking for your domain with the
command:
sudo vdsClient -s 0 getStorageDomainInfo
d7b9d7cc-f7d6-43c7-ae13-e720951657c9
4. (VDSM HOST) Activate all LVs for the storage domain
sudo lvm lvchange --config 'global {use_lvmetad=0}' -aay
d7b9d7cc-f7d6-43c7-ae13-e720951657c9
... Repeat the following steps for each image you want to recover ...
5. (VDSM HOST) Get the list of volumes in the image
sudo vdsClient -s 0 getVolumesList
d7b9d7cc-f7d6-43c7-ae13-e720951657c9 <old-SP> <imageID from step 1>
6. (VDSM HOST) Determine the LEAF volume (if the image has only ony volume, it is the
leaf)
sudo vdsClient 0 getVolumeInfo d7b9d7cc-f7d6-43c7-ae13-e720951657c9
<old-SP> <imageID> <volumeID>
Look for 'voltype = LEAF'
7. (VDSM HOST) Copy the volume to a new file
sudo qemu-img convert -p -O raw
/dev/d7b9d7cc-f7d6-43c7-ae13-e720951657c9/<volID> <volID>.new
I am right with you, and I don't understand why nobody else from
your colleagues doesn't suggest anything...
It's weird.
Anyway thanks for your help.
Regards
Alain
Alain VONDRA
Chargé d'exploitation des Systèmes d'Information Direction
Administrative et Financière
+33 1 44 39 77 76
UNICEF France
3 rue Duguay Trouin 75006 PARIS
www.unicef.fr
-----Message d'origine-----
De : Adam Litke [mailto:alitke@redhat.com] Envoyé : jeudi 9 avril 2015
19:35 À : VONDRA Alain Cc : users(a)ovirt.org Objet : Re: [ovirt-users]
Storage domain not in pool issue
On 09/04/15 14:16 +0000, VONDRA Alain wrote:
>Hi,
>What do you about my last email, do you have a non destructive option to solve my
problem ?
The approach I suggest would leave the original data on the broken storage domain so
it's non-destructive. That's really all I am able to offer as a path forward. I
think we've given others ample time to respond.
--
Adam Litke
--
Adam Litke