My backbone network is 10GB using 9000b jumbo frames. Today I tried to change my TrueNas scale (ElectricEel-24.10.2.1) settings to optimal support that. … I did not manage.
The NAS is connected to the network using a ConnectX4-card (one port used with a 10G SFP+). That 10Gbe port is used as a trunk towards the network and is carrying all the VLAN’s.
My idea is to leave the mtu of all low traffic vlans at 1500 and the vlans carrying heavy load to 9000.
MTU’s are a bit vague term. It can refer to the content and e.g to the L2-frame.
The L2-mtu should be bigger than the content MTU since it has overhead like vlan tags and more.
So for the L2-MTU you see values like 9014 (default connectx4) and higher values e.g. 9200.
And I do not know if the value as to be entered at the interface level should have the content value (9000) or the L2-value (higher 9014??).
Whatever I decide to define 9014 for the trunk. No problem. System still running and accessible.
Than I did go to the TrueNas base VLAN, I mean the VLAN having the core NAS stuff: default route, the gui, the SMB and iSCSI etc.
That is the vlan carrying the heavy traffic, so I changed the MTU to 9000 since I regard the vlan as the endpoint (I am not using a bridge here, perhaps I will try that as well later)
The result … the GUI was not accessible any more. So I had a big big issue.
After some time I did discover that the interface settings are changeable via a bootmenu option. With that option I could make the system accessible
and I could try other values for the vlan MTU.
The result of those tests where … simple … all MTU values <> 1500 do make the GUI-inaccessable …
So I really wonder how to use jumbo frames withour loosing the GUI !!
Note here a related but other issue. There is only one routing table with one default route. Where I really, really would love a default route per vlan (vlans as related to the NAS-core. I am also using vlans’s for my VM’s they are routing per vlan like it IMHO should)