Force vcenter to clone by a vmnic

[jucagol]

I have two servers connected to a switch at 1gb to reach internet and another one at 10gb to Inter-Connect both servers.

when I try  to clone virtual machines with vcenter always uses the 1gb connection. Both ports, are connected to the same vswitch, but still uses the 1gb instead of the 10gb.

Is there any way to Force the clone to be done by the 10gb adapter?

thanks!

[Kinnison · Accepted solution]

Hi,

If memory serves me correctly, cloning is a "provisioning" type traffic (allow me the expression) which uses the default VMKernel and TCP/IP stacks. You could set up a dedicated provisioning network with its own VMKernel and "Provisioning TCP/IP" stack and thus using a VMNic of your choice, up to you, right or wrong I personally do that.

Regards,
Ferdinando

[Lalegre]

Hello @jucagol,

As @Kinnison mentioned, he is correct, you need to create a VMkernel with Provisioning checkbox into it and make sure it is connected to a portgroup where the 10Gbps link is being used.

[Kinnison]

Hi,

Excuse me but I forgot to mention a "detail" that could be by no means negligible that you find documented here:
https://core.vmware.com/resource/vsphere-vmotion-unified-data-transport

Honestly, I'm not used to cloning virtual machines involving different HOSTs with local storage to the same ones (generally I don't use "clones" at all), however in cases of cold migration the difference "is there and we can see it".

Regards,
Ferdinando

[jucagol]

Thanks for all the comments. 

I apologize because I forgot to describe the scenario deeply. One of the servers is 6.7.0 version and the other one is 7.0.3. the vcenter is located at the newest server (7.0.3).

I´ve created the vmkernel inside a new portgroup connected to a vswitch connected to a vmnic at 10Gb, I´ve already done the same on the other Host.

The first test to clone failed due to a host disconnection, and I´ve repeated twice with the same result. I wonder if I need to move the virtual machine to the same vswitch, same port group as new vmkernel created.

On the other hand, I´ve set an address 10.0.0.x (.253 and .254 on each side), when I try to ping to the othe host in this address space I don´t receive any response from the other side.

I´ve unmark in the 1GB vmkernel all the setting except maintenance check. I´ve selected all the other functions in the new vmkernels created connected at 10GB.

I think I left some settings not properly configured.

[jucagol]

After change the 10G adapter to an empty vswitch and the new vmkernel to the same vswitch the cloning is working at 10Gb, so the problem was to share the vswitch with both vmkernel and vmnic (1GB and 10Gb). once moved to another vswitch and leave it alone, the cloning works fine.

Thanks, your answer clarified my mind and testing your configuration worked fine.

[jucagol]

Update: I thougth than cloning worked fine, but when I monitored the traffic and I realized the speed is really slow, in fact, the adapte me is ready to supoort 10g but the speed is lower than 100mbps, I’ve already review the config but looks everything fine.

[Kinnison]

Hello,

Comment edited to remove its content as it no longer serves the context of this discussion.

Regards.

[jucagol]

I still have the same issue, I cannot clone virtual machines from one node to another with the 10gb network link. I have to move VM with 900gb of hard disk and it takes up to 48hours to be cloned because the highest speed in the link is 20Mbps. We cannot have th customer without systems more than 48 hours, this is the reason why we need the real network link speed.

Anyone can help me?

Thank in advance.

Reminder, 2 esxi servers stand alone connected to a 10Gb switch, and the vmk adapter is assigned to this nics.

[Kinnison]

Hello,

Sorry but in my opinion there is something that doesn't add up to me, I just finished cloning a powered off virtual machine with disk drives for a total of 480 Gigabyte (thick provisioned eager zero) from one host to another (local storage on both) "via the network" using a gigabit link and took roughly 65 minutes. So, what I don't understand is how you determined that in your case it takes more than 48 hours.

Well, are you sure you don't have some problem with the storage subsystem, which plays a substantial role, even before other possible network problems. If I were in your place I would make sure of it.

Regards,
Ferdinando