VM MTU

[abarakat]

Is their a way to change MTU on the VM interface without hotpluging that MTU into the OS?

Reason for this is that I want the VM hardware interface to be ready for jumbo frames but I want to be able to keep VM O/S default 1500 MTU and change it as needed @ the os level back and forth without needed to keep changing this @ the VM hardware interface level.

Thanks

 

[VictorSTS]

You can set MTU for each NIC in hardware settings of the VM. Tick the checkbox "advanced" and set MTU. You will need to force a (lower) MTU in the guest OS, as most tend to use the MTU of the hardware.

 

[abarakat]

You can set MTU for each NIC in hardware settings of the VM. Tick the checkbox "advanced" and set MTU. You will need to force a (lower) MTU in the guest OS, as most tend to use the MTU of the hardware.

Thank you but is their a way to stop the software from adopting the virtual hardwares MTU? We want to keep the hardware flexible with high MTU but the software to it's defaults.

 

[VictorSTS]

Sorry, I don't get what you are trying to accomplish, please elaborate.

 

[abarakat]

Sorry, I don't get what you are trying to accomplish, please elaborate.

We do software/Hardware development lots of testing as such. In VMware we can set the portgroup to any MTU we want. This does not affect the VM. vm interface can be set to any MTU as needed by users. Most users do not need to change MTU on the VM so no harm its stays at its default of 1500. When they need to test diffrent MTU they just change it in the VM as they please and not have to worry about anything after that.

With Proxmox if you configure the hardware MTU it will change the OS level MTU this is disruptive. affecting the default MTU that most users need to test with. Also we have seen OS crashes when proxmox tries to hotplug the hardware set MTU on the unsupported OS.

I just want to be able to set MTU on the VM hardware side without it trying to set it at the VM os level. I dont want it to hotplug in the MTU to the OS.

 

[abarakat]

Sorry, I don't get what you are trying to accomplish, please elaborate.

Nevermind I figured out how to disable hotplug but this is not what I was looking for. I want to be able to hotplug on the VM hardware side but not affect the VM OS side. Seems that this may not be possible.

 

[VictorSTS]

Cant really compare with VMWare here, I don't know how your use case is implemented on ESXi nor how it works internally...

On PVE, set the physical NIC / bond / bridge MTU to the highest possible MTU you want to support (and that your switches support, of course).
On each VM, virtual NIC will use 1500 by default unless you set a different value. These values could be as high as the MTU set in the physical NIC / bond / bridge. You can have 9000 MTU in physical NIC / bond / bridge and just 1500 in the VMs.
On the guest OS, you can set any MTU you want up to the value set in that virtual NIC. I you want the guest OS to use a given MTU instead of the virtual NIC MTU's, tell it to do so as AFAIK guest OS will use the virtual NIC MTU by default. IMHO, it would be easier/safer for the VM users to be able to change the MTU setting in the guest OS instead of having to deal with VM settings.

A different issue is that the NIC hotplug that happens when you change MTU, doesn't work for your guest OS. Has always work for me in a miriad of Windows and Linux OS. Feel free to open a thread, maybe someone can help with that.