• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi vSwitch MTU

Adam12176

Limp Gawd
Joined
Dec 27, 2007
Messages
205
Hey all. I've run into a bit of an odd problem with ESXi 4.0u1.

Background: I'm messing around with ESX/iSCSI in preparation for an MD3000i in house. I'm currently using OpenFiler, and a Win2k8R2 machine for various testing/proof of concept.

The solution works in a bare bones configuration currently. I set up a separate vSwitch - at first, not even tied to a physical adapter, addressed properly, and I was able to map LUNs on the Windows machine.

The problem is jumbo frames. It appears that ESX supports jumbo frames, albeit appearing to be unsupported officially. Most documents reference enabling jumbo frames on the vmkernel however. Right now, after issuing the command 'esxcfg-vswitch -m 9000 vSwitch2' (iSCSI vSwitch) the two will not utilize jumbo frames. The R2 box is configured to use jumbo frames on the 'iSCSI' interface, as is OpenFiler. A 9k frame ping fails to respond in either direction.

The twist: SSH'ed into the ESX box, it WILL ping the VMkernel port with a 9K frame.

I have since associated the iSCSI vSwitch with a physical adapter, thinking that may have been causing the problem. No dice. I feel like the ESX config is solid, as it will ping itself with the appropriate frame size.

Again, this is just proof of concept, I'm doing this more for my own benefit, as the real solution will have appropriate physical hardware. I'm more mystified as to why this isn't working at this point. Any comments/suggestions are appreciated.
 
ESX4 supports JF for storage, 3.5 didn't ;)

In 4, you have to set the switch, the vmkernel port, and hte physical port to 9000 MTU. Let me check how to do all of that on i, as I'm far more used to classic.
 
are you using VMKping? You specifically have to add change the adapter and switch MTU to enable jumbo frames.

# esxcfg-vmknic --server <servername> -a -i 10.10.10.10 -n 255.255.255.0 -m 9000 iSCSI2
 
are you using VMKping? You specifically have to add change the adapter and switch MTU to enable jumbo frames.

# esxcfg-vmknic --server <servername> -a -i 10.10.10.10 -n 255.255.255.0 -m 9000 iSCSI2

But I'm not using the kernel to pass traffic. I have my 'dumb' vSwitch set to MTU 9000, and all I want that switch to do is pass traffic from the OF box to the R2 box. No kernel involvement.
 
You need to make both of the virtual adapters support jumbo frames, then enable jumbo within the respective VM guests. You may have the best luck by replacing with a vmxnet2 or vmxnet3 adapter, not the generic e1000 adapter.
 
Quick update - changed the adapter for both the R2 and OF boxes to vmxnet 3 and vmxnet 2 enhanced respectively, and it works.

Thanks archival, I didn't think the adapter type would prohibit JF from working in this case, especially because the Windows control panel had the option listed as one would expect.

As an edit however - the transfer rates on a single large file are ri-donkulously slow. A little surprised at that fact. It is just a SATA disk, but I expected a little higher than 2-5Mb/sec. Proof of concept, it works. Just a bit peculiar.
 
Last edited:
That's odd. Did you try without JF?

(reality - payload overhead is so small that JF only buy you a percentage point or two most of the time, assuming it works. And there are chances that it will make things far worse if it doesn't work).
 
Back
Top