Non-coherent vds/switch MTU config can break backups

[fcambot]

Hi,
here is a bad trip i had i want to share.

I changed the 48 ports switch i use in my production stack (i had PoE, not needed, swapped with a non-PoE).
I did not remember immediatly that the replaced switch had jumbo frames (MTU 9000) enabled, and i did not check the new one. All my vsphere virtual switchs have a 9000 MTU.

The nigth next to replacement, every job failed with an NFC error can not get next FILE_MESSAGE or so, crashing at "getting info from vsphere" point.
Everything else was functionning (browsing VM, recalculate size,etc).

Once i enabled jumbo frames on the switch, every thing went fine.

Additional not : i first tried to set the MTU to 1500 on the management virtual switch of a vSphere, and it turned ugly : i can't anymore connect by ssl or gui, and now i have to choose between restore default vsphere config or perform a full reinstall.
The network restore config options failed.

[PTide]

Hi,

The behaviour described is expected and there exists an article with the following said:

To allow an ESX host to send larger frames out onto the physical network, the network must support Jumbo Frames end to end.
Ensure that your NIC or LOM supports Jumbo Frames.

and now i have to choose between restore default vsphere config or perform a full reinstall.

So, do you have problems now or turning on jumbos on switch has resolved the issue?

Thank you.

[fcambot]

No backup problems anymore