Re: Backup VM with hardware passthrough

Sorry for reviving an old post but is backup possible with a USB passthrough on XCP-NG 8.3? or at least a snap shot?

I had to just play around with it and see what happens, I feel like I watched videos explaining it fairly good but I you are right. Theres tons of options and they keep adding more which is nice. Just makes understanding whats happening a little more challenging. If you have hardware for it spin up a test lab and try it out.

@CodeMercenary I use the paid for XOA and it backups up its config to the cloud so it wouldn't be a problem for me to restore. You can easily set something like that up by backing up to a store location then having that backup to a cloud storage location.

As far as having the XO instances running Im not sure how that would work. I know with NBD storage you should be able to connect to the hard drive with multiple XO's (I think) but the problem I ran into quickly was the fact that when I made a backup using my XOA, the XO from source would see it as a backup without an associated job. If you had three XO's running with 3 different remote storage locations I don't think you'd have a problem with that. Instead you might run into the XO's seeing the 3 different backup jobs running , 2 of which, they don't know about.

I use XOA to backup to a different storage array, then I have that storage backup to the cloud. My backups aren't really up to "standard" but thats all I can do for now. My plan is to make backups to a different storage pool on the same central storage I run my VM's on, back that whole unit up to an off site unit with ZFS replication, then back that backup to the cloud with StorJ. Right now Ive just got the one backup on the same Storage server and the StorJ.

I do this but its not so that I can have an exact copy of the XOA. I just use it in the rare event I want to modify my XOA like give it more core or more ram. Its easy enough to spin up a community edition XO and then use it to add remove resources, Also I keep my XOA and my XO from source on different host in case one goes down. Nice easy lazy way of still being able to manage my VM's

I don't know of anything to keep it from deactivating but usually Microsoft will let you reactivate if you call thier support line. This has happened to me several times. Ive found they will let you reactivate at least three times. IDK if thats still the case though. This was years ago

I would run it in Scale if i were you. Ive read about people's set ups like this and there hasn't been any issues with it but It has been argued that virtualization of your storage underneath a hypervisor is dangerous. I don't know what type of pitfalls you might encounter with this but it will surely make recovery from a disaster a little more complicated. And with Scale being able to run VM's you can have the GPU passed through to your Win11 machine.

I presume you are transcoding in your Jellyfin? I use Plex and I pass through a GPU to it (but thats in esxi). Can you pass through your GPU to multiple devices to use for that? I know there is vGPU but I don't know if it can be used for transcoding. It would be interesting to try though

I read a forum post about some of the xo community editions having issues with that. Something to do with invalid certificates or something. Did it ever fail or did it just stay at 0 forever?

Oh, I didn't know you tried using virtual box to do an export. I haven't tried that before.

I cdon't know about a guide anywhere to set it up but I can say that there is definitely nothing wrong with running two XO instances. I have XOA and an XO community edition running in the same pool. I always make sure if I am about to do a big change that my XOA and my XOCE are on different hosts. And tbh its mostly because Im not as comfortable in the CLI with xe commands. But there were at least two times where this came in handy. Especially during my initial build stage as I broke everything twice. the XOCE saved me from having to try and start over on my XOA.

As far as I know you can't "sync" them. You can however pull a config off one and restore it to another one. You just have to remember to do that anytime you make changes to XO. Also you don't want two XO's with backup jobs enabled. Another downside is that if you give them both all the same resources it will look like there are errors in your backups. I vaguely remember it saying "Backup for VM found but no associated job was found" or something. Innocuous error but annoying none the less.

How did you export it? I used the Vmware export tool on Esxi 6.5 and none of those exports worked for me. I ended up having to use clonezilla and it was able to make an OVA that worked.

@splastunov Dom0 would need to have a vGPU in this scenario?

@splastunov Do the AMD GPU's not require a license?

Its almost certainly goes through XOA. When I was moving from Esxi to XCP-NG I abandoned the process, then beefed up my XOA memory and CPU cores and it sped up marginally. That was a while ago, some of the VM's I had to use Clonezilla on (too large and kept failing). From what Ive read the migration tool has improved dramatically since then.

If you have hardware that supports this you can try to do an openvpn Server as a TAP adapter. it will be closer to a Layer 2 MPLS connection than an IPSEC tunnel. I have no idea if XCP-NG or XO/XOA use any type of Multicast traffic to communicate. Thats the only thing I can think of unless you have IPSEC firewall policies blocking some type of traffic that XO needs. I presume you can ping/ssh into the remote host? Are these hosts using shared storage over the MPLS/IPSEC?

Edit:
I just found this as well.
https://xcp-ng.org/forum/topic/6609/unable-to-join-a-host-to-a-pool

Seems to be also that the XCP-NG hosts will attempt to talk to each other over local URL's in some cases but I can't find a post about that. Bridging the two networks the same way the MPLS was would likely get you back to the way it was.

Out of curiosity what are you using this for? what are these VM's doing?

So I bet you don't have to use a heater in the winter lol.

While migrating from Vmware to XCP-NG I ran into an issue where Clonezilla wouldn't load after the initial boot menu when using the Windows Server 2012 R2 template. I received this error:

Timed out for waiting the udev queue being empty

Fortunately I just ran into a similar issue where Clonezilla would have a scrambled display after the boot menu with the Windows 10 Template and I found this forum post:

Re: Scrambled display booting Clonezilla (and linux) ISOs under Win10 template

using this command:
xe vm-param-set uuid=<UUID of your VM goes here, no brackets> platform:device_id=0001

Then booting CloneZilla and selecting the KMS boot option fixed the issue. After cloning shut down the VM and change the device ID back with:

xe vm-param-set uuid=<UUID of your VM goes here, no brackets> platform:device_id=0002

I would have replied on the original post but it wasn't the same issue and the post was a little old. Hope this helps