On Thu, Jan 17, 2019 at 4:53 PM Gianluca Cecchi <gianluca.cecchi(a)gmail.com&gt; wrote: you can export them to ova on some device that can later be mounted to the destination environment. this is similar to the export domain functionality - but you didn't specify why the export domain functionality is not applicable for you. in this case, you can optimize the above with [1] - export the vm as ova on one of the hosts that is installed with the ovirt-sdk and upload the ova from that host to the destination environment. I don't think it is possible - you'll need the VM to be configured but not running on the host (in this case, the engine asks vdsm to destroy the vm). and anyway I wouldn't do that - it is targeted for importing from foreign environments. one of the side effects is that you'll lose the oVirt-specific configuration of the vm. Or is there (eventually via command line) a way to use virt-v2v in this virt-v2v is just an overkill for this scenario. [1] https://github.com/oVirt/ovirt-engine-sdk/blob/master/sdk/examples/upload...

On Thu, Jan 17, 2019 at 4:47 PM Gianluca Cecchi <gianluca.cecchi(a)gmail.com&gt; wrote: BTW: would it be possible to export as ova on a path on host that is an NFS share, bypassing export domain setup, or does it require local storage? Gianluca

On Thu, Jan 17, 2019 at 6:54 PM Gianluca Cecchi <gianluca.cecchi(a)gmail.com&gt; wrote: Right The first one is included in 4.3 already (in general, the ova handling is better in 4.3 compared to 4.2 in terms of speed). The second is unlikely to happen as we found there's no real need for it. Yes, but note that you need to mount the NFS share in a way that the root user on host can write to the specified path. Unfortunately, it is indeed done that way in 4.2 :/ Let me explain: We wanted disks within the ova to be of type qcow (for several reasons, like thin-provisioning, compression). In 4.2 we create temporary disks on the storage that are then copied into the OVA (and then removed). In 4.3 qemu-img converts the images directly into the place they should be within the OVA. That's the reason for the export process being faster in 4.3.

On Thu, Jan 17, 2019 at 6:57 PM Arik Hadas <ahadas(a)redhat.com&gt; wrote: Tried by I got error. The VM from which I try to create OVA is composed by 3 disks: 15 + 60 + 440 Gb This is the sequence of events seen in engine: Starting to export Vm dbatest5 as a Virtual Appliance 1/17/19 5:33:35 PM VDSM ov200 command TeardownImageVDS failed: Cannot deactivate Logical Volume: ('General Storage Exception: ("5 [] [\' Logical volume fa33df49-b09d-4f86-9719-ede649542c21/08abaac5-ef82-4755-adc5-7341ce1cde33 in use.\']\\nfa33df49-b09d-4f86-9719-ede649542c21/[\'08abaac5-ef82-4755-adc5-7341ce1cde33\']",)',) 1/17/19 9:48:02 PM Failed to export Vm dbatest5 as a Virtual Appliance to path /export/ovirt/dbatest5.ova on Host ov200 1/17/19 9:48:03 PM Disk dbatest5_Disk1 was successfully removed from domain ovsd3750 (User admin@internal-authz). 1/17/19 9:48:04 PM Disk dbatest5_Disk2 was successfully removed from domain ovsd3750 (User admin@internal-authz). 1/17/19 9:48:05 PM Disk dbatest5_Disk3 was successfully removed from domain ovsd3750 (User admin@internal-authz). 1/17/1 99:48:05 PM And this left this file [root@ov200 ~]# ll /export/ovirt/dbatest5.ova.tmp -rw-r--r--. 1 root root 552574404608 Jan 17 22:47 /export/ovirt/dbatest5.ova.tmp [root@ov200 ~]# The ".tmp" extension worried me about possibly not completed ova... is this the case... anyway I tried then to import it, see below I have not understtod which LV it tries to deactivate... On destination host I get this, but I don't know if it depends on ova not exactly completed in its part; from the "Broken pipe" error I suspect so...: ./upload_ova_as_vm.py /export/ovirt/dbatest5.ova.tmp RHVDBA rhvsd3720 Uploaded 69.46% Uploaded 69.70% Uploaded 69.95% Uploaded 70.21% Uploaded 70.45% Uploaded 70.71% Uploaded 70.72% Traceback (most recent call last): File "./upload_ova_as_vm.py", line 227, in <module> proxy_connection.send(chunk) File "/usr/lib64/python2.7/httplib.py", line 857, in send self.sock.sendall(data) File "/usr/lib64/python2.7/ssl.py", line 744, in sendall v = self.send(data[count:]) File "/usr/lib64/python2.7/ssl.py", line 710, in send v = self._sslobj.write(data) socket.error: [Errno 32] Broken pipe [root@rhvh200 ~]# I have verified on a 4.3rc2 env that I can indeed execute "export as ova" for a running VM too. I have a CentOS Atomic 7 VM and when you export as ova, a snapshot is executed and the the ova file seems directly generated: [root@hcinode1 vdsm]# ll /export/ total 1141632 -rw-------. 1 root root 1401305088 Jan 18 11:10 c7atomic1.ova.tmp [root@hcinode1 vdsm]# ll /export/ total 1356700 -rw-------. 1 root root 1401305088 Jan 18 11:10 c7atomic1.ova [root@hcinode1 vdsm]# And at the end the snaphsot has been correctly removed. Vm c7atomic1 was exported successfully as a Virtual Appliance to path /export/c7atomic1.ova on Host hcinode1 1/18/19 11:10:23 AM Starting to export Vm c7atomic1 as a Virtual Appliance 1/18/19 11:08:47 AM OK. Ok, verified that without the no_root_squash I was not able to start the export phase That way I think you should do at least some kind of estimation before starting, otherwise yo can compromise primary storage of the infrastructure... In my case I have to export about 500Gb of disks and I risked to fill up the storage domain ;-( Verified this in 4.3 RC2 Wouldn't be simpler to do it the vSphere way, where you cannot export as ova IIRC (but you can import from ova), while you can export to ovf? This way you can directly export the disks and the VM definition on target directory Anyway if 4.3 solves this, it could be ok Do you have any hint about errors on 4.2 or any information I could provide? Gianluca

On 1/17/19 8:49 AM, Gianluca Cecchi wrote: The only way I've ever done this is via an Export Domain.