On Fri, Nov 23, 2018 at 1:12 PM AG <agmeshs@gmail.com> wrote:

Hello,

Our cluster use NFS storage (on each node as data, and on a NAS for export)

Looks like you invented your own hyperconverge solution.

The NFS server on the node will always be accessible on the node, but if it is

not accessible from other hosts, the entire DC may go down.

Also you probably don't have any redundancy, so failure of single node cause

downtime and data loss.

Did you consider hyperconverge Gluster based setup?

https://ovirt.org/documentation/gluster-hyperconverged/Gluster_Hyperconverged_Guide/

We currently have a huge problem on our cluster, and export VM as OVA
takes lot's of time.

The huge problem is slow export to OVA or there is another problem?

I checked on the NFS storage, and it looks like there's files who might
be OVA in the images directory,

I don't know the ova export code, but I'm sure it does not save in the images

directory. It probably creates temporary volume for preparing and storing the

exported ova file.

Arik, how do you suggest to debug slow export to OVA?

and OVF files in the the master/vms

OVF should be stored in OVF_STORE disks. Maybe you see files

created by an old version?

Can someone tell me if, when i install a new engine,
there's a way to get this VMs back inside the new engine (like the
import tools by example)

What do yo mean by "get this VMs back"?

If you mean importing all vms and disks on storage to a new engine,

yest, it should work. This is the base for oVirt DR support.

ps : it should be said in the documentation to NEVER use backup of an
engine when he is in a NFS storage domain on a node. It looks like it's
working, but all the data are unsynch with reality.

Do you mean hosted engine stored on NFS storage domain served by

one of the nodes?

Can you give more details on this problem?

Please also specify oVirt version you use.

Nir