[SOLVED] Advice needed - Migration from ESXi to PVE - how hard can this be?

[zzz09700]

Here's the situation in front of me:
1. We have several old, not-so-powerful Broadwell, Haswell and IvyBridge servers (yeah I know its like a decade old) running ESXi 6.5 and 7, each with a handful of VMs on local disk.
2. After some evaluations we found that two single socket EPYC servers are more than capable of handling all our vms, so went ahead and brought those.
3. The old ESXi servers were not clustered in any way and we intend to keep this configuration - the two EPYC servers will be running as two standalone servers.
4. Some of our VMs has USB expansion card or GPU passed through to it (and that's why we never bothered with applying HA related stuff)

Since the free ESXi is history we are evaluating the option of installing PVE on the EPYC servers, and I'm wondering how hard it is to import ESXi VMs into PVE, with all the passthrough and stuff still in place.

We will be doing a complete offline impot, which means old ESXi servers are shut down, all VMs exported to other server's disks and some of the NVME disks on old ESXi servers will be wiped and then installed into EPYC servers. Here downtime related stuff is not a concern.

What I need to ask:
1. Is it possible to import ESXi-exported ovf and vmdk into PVE? We can reconfigure passthrough manually later - the GPUs has to be moved by hand anyway.
2. If it is possible, can we do that in an one-click fashion, within the web-gui of PVE, or is there any guide on how to do it with command line? We do have devs working on Debian/Ubuntu workstations all day long but none of us know how to tinker with KVM or qemu related stuff.

 

[cheiss]

Hi,

we do have quite extensive guide for (esp. from ESXi) migration to Proxmox VE, which I'd firstly recommend: https://pve.proxmox.com/wiki/Migrate_to_Proxmox_VE

1. Is it possible to import ESXi-exported ovf and vmdk into PVE? We can reconfigure passthrough manually later - the GPUs has to be moved by hand anyway.

OVF/VMDKs can be imported directly, which is laid out in detail in the Import OVF section in the guide above.

2. If it is possible, can we do that in an one-click fashion, within the web-gui of PVE, or is there any guide on how to do it with command line? We do have devs working on Debian/Ubuntu workstations all day long but none of us know how to tinker with KVM or qemu related stuff.

We do have an web-based import wizard for exactly that, documented in the Automatic Import of Full VM, again in the guide.

As for USB and PCI(e) passthrough, that has to be configured afterwards.
- USB should be quite simply using the web UI, under Hardware > Add > USB Device.
- PCI passthrough is unfortunately a bit more complicated, but we again have pretty extensive documentation for that too: https://pve.proxmox.com/pve-docs/chapter-qm.html#qm_pci_passthrough and https://pve.proxmox.com/wiki/PCI_Passthrough

 

[zzz09700]

Hi,

we do have quite extensive guide for (esp. from ESXi) migration to Proxmox VE, which I'd firstly recommend: https://pve.proxmox.com/wiki/Migrate_to_Proxmox_VE



OVF/VMDKs can be imported directly, which is laid out in detail in the Import OVF section in the guide above.

Hi there, thanks for the fast reply.

I'm reading through that section and kinda get lost in the process.
What we get from ESXi VM export is a ovf file, a md file and a vmdk file. AFAIK ovf contains metadata of the VM and vmdk is the actual disk image. So when we do `qm importovf`, are we importing a VM without any disk image? It seems the disk image is going to be converted to qcow2 and attached to the VM in a seperate command `qm disk import`, am I getting this right?

 

[cheiss]

So when we do `qm importovf`, are we importing a VM without any disk image?

No, the qm importovf will also import (and possibly convert as needed) the disk images.

There is the following note in the Import OVF section:

Once exported, navigate to the export directory that contains the {VM name}.ovf and matching vmdk file.

I.e. both the .ovf file as well as any .vmdk file produced by ovftool must be accessible by Proxmox VE.

The exact syntax of that command is:

qm importovf {vmid} {VM name}.ovf {target storage}

So it will create a new VM with the specified VM-ID, taking the configuration from the .ovf metadata file and importing all disks that belong to it to the specified target storage.

Hope that clears it up! Feel free to ask if something else is unclear!

 

[zzz09700]

OK got it. But since we are not going to use ovftool, I'm wondering if we need to change the name of vmdk file manually. What we get from ESXi VM export is something like, for example, {vm-name}.ovf and {vm-name}-1.vmdk. Does that vmdk with `-1` postfix count as `matching vmdk file`?

 

[cheiss]

I'm wondering if we need to change the name of vmdk file manually. What we get from ESXi VM export is something like, for example, {vm-name}.ovf and {vm-name}-1.vmdk. Does that vmdk with `-1` postfix count as `matching vmdk file`?

Yes, that is expected. You can inspect the referenced disk images when opening the .ovf files, under the <references> block if IIRC.
So the import into Proxmox VE should work without any problems.

But since we are not going to use ovftool

I guess that means you are simply exporting them via the web interface? The ovftool works the same AFAIK, just gives you a bit more flexibility in where to store the files.

 

[zzz09700]

That clears everything for us. We will be spinning up the EPYC servers and see how PVE performs soon.

Thanks again for the awesome community support. The level of community support of PVE is even better than what we got from vmware paid phone support.