[ovirt-users] Disks Illegal State

[Clint Boggio]

In my case Markus, the backing disks are MIA and show only as bright
red broken symbolic links. Using the postgres commands to set them as
OK would be folly, and likley cause more trouble. if the snapshot disks
are truly gone, (and they are), what procedure would i use to inform
the database and set the VM's in a usable status status again ?

On Mon, 2016-04-18 at 12:39 +0000, Markus Stockhausen wrote:
> > 
> > Von: users-bounces at ovirt.org [users-bounces at ovirt.org]" im
> > Auftrag von "Clint Boggio [clint at theboggios.com]
> > Gesendet: Montag, 18. April 2016 14:16
> > An: users at ovirt.org
> > Betreff: [ovirt-users] Disks Illegal State
> > 
> > OVirt 3.6, 4 node cluster with dedicated engine. Main storage
> > domain is iscsi, ISO and Export domains are NFS.
> > 
> > Several of my VM snapshot disks show to be in an "illegal state".
> > The system will not allow me to manipulate the snapshots in any
> > way, nor clone the active system, or create a new snapshot.
> > 
> > In the logs I see that the system complains about not being able to
> > "get volume size for xxx", and also that the system appears to
> > believe that the image is "locked" and is currently in the snapshot
> > process.
> > 
> > Of the VM's with this status, one rebooted and was lost due to
> > "cannot get volume size for domain xxx".
> > 
> > I fear that in this current condition, should any of the other
> > machine reboot, they too will be lost.
> > 
> > How can I troubleshoot this problem further, and hopefully
> > alleviate the condition ?
> > 
> > Thank you for your help.
> Hi Clint,
> 
> for us the problem always boils down to the following steps. Might be
> simpler as we use
> NFS for all of our domains and have direct access to the image files.
> 
> 1) Check if snapshot disks are currently used. Capture the qemu
> command line with a "ps -ef"
> on the nodes. There you can see what images qemu is started with. For
> each of the files check
> the backing chain:
> 
> # qemu-img info /rhev/.../bbd05dd8-c3bf-4d15-9317-73040e04abae
> image: bbd05dd8-c3bf-4d15-9317-73040e04abae
> file format: qcow2
> virtual size: 50G (53687091200 bytes)
> disk size: 133M
> cluster_size: 65536
> backing file: ../f8ebfb39-2ac6-4b87-b193-4204d1854edc/595b95f4-ce1a-
> 4298-bd27-3f6745ae4e4c
> backing file format: raw
> Format specific information:
>     compat: 0.10
> 
> # qemu-img info .../595b95f4-ce1a-4298-bd27-3f6745ae4e4c (see above)
> ...
> 
> I don't know how you can accomplish this on ISCSI (and LVM based
> images inside iirc). We
> usually follow the backing chain and test if all the files exist and
> are linked correctly. Especially
> if everything matches the OVirt GUI. I guess this is the most
> important part for you.
> 
> 2) In most of our cases everything is fine and only the OVirt
> database is wrong. So we fix it
> at our own risk. Because of your explanation I do not recommend that
> for you. It is just for 
> documentation purpose.
> 
> engine# su - postgres
> > 
> > psql engine postgres
> > 
> > select image_group_id,imagestatus from images where imagestatus =4;
> > ... list of illegal images
> > update images set imagestatus =1 where imagestatus = 4 and <other
> > criteria>;
> > commit
> > 
> > select description,status from snapshots where status <> 'OK';
> > ... list of locked snapshots
> > update snapshots set status = 'OK' where status <> 'OK' and <other
> > criteria>;
> > commit
> > 
> > \q
> Restart engine and everything should be in sync again. 
> 
> Best regards.
> 
> Markus