Bridge does not inherit MTU from ports in PVE 7 using ifupdown2

[gartoffel]

Consider the following example config in /etc/network/interfaces:

auto lo
iface lo inet loopback

auto mgmt0
iface mgmt0

auto mgmt1
iface mgmt1

auto north0
iface north0
    mtu 9000

auto north1
iface north1
    mtu 9000

auto bond0
iface bond0
    bond-slaves north0 north1
    bond-miimon 100
    bond-mode 802.3ad
    bond-xmit-hash-policy layer3+4
    bond-lacp-rate 1
    mtu 9000

# Bridge for north interfaces
auto vmbr0
iface vmbr0
    bridge-ports bond0
    bridge-stp off
    bridge-fd 0
    bridge-vlan-aware yes
    bridge-vids 301
    mtu 9000

When issuing ifreload -a -s (which runs a syntax check of /etc/network/interfaces) I receive the following warning:

warning: vmbr0: bridge inherits mtu from its ports. There is no need to assign mtu on a bridge

Let's quickly verify that the MTU of the bridge and the interfaces is correct, though:

root@proxmox:~# ip a show dev bond0 | grep mtu
8: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 9000 qdisc noqueue master vmbr0 state UP group default qlen 1000

root@proxmox:~# ip a show dev vmbr0 | grep mtu
9: vmbr0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 qdisc noqueue state UP group default qlen 1000

Okay, then let's remove the mtu setting from vmbr0, run ifreload -a -s again (this time without getting a warning) followed by ifreload -a and see what happened:

root@proxmox:~# ip a show dev bond0 | grep mtu
8: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 9000 qdisc noqueue master vmbr0 state UP group default qlen 1000

root@proxmox:~# ip a show dev vmbr0 | grep mtu
9: vmbr0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000

That's unfortunate. Is that a bug in ifupdown2 that needs to be addressed?

 

[spirit]

Hi,
thanks for reporting. I'll do test on my side to verify that and contact ifupdown2 devs if I can reproduce.

(I recommand to always set mtu everywhere to be sure, even if it's display this warning)

 

[auranext]

Hi,

I ve noticed same behavior on PVE7.0 nodes
previous versions pve6 and pve5 running with ifupdown does not have the pb
@spirit ,what conclusions did you make ?

 

[iniciaw]

I believe than is only and alert, If you use MTU 9000 on bond0 you don't need add on bridge, because bridge get config from bond0.
This alert only give information, don't force leave MTU config from bridge, only inform that it is not required

 

[auranext]

@iniciaw what you describe is the expected behavior, and I have not notice any issue with bond if.
For bridges interfaces PVE6 and PVE7 have different behavior.
let me clarify this
With same /etc/network/interfaces file , only physical nics and dummys nic have mtu 8950 specified
On PVE6 (ifupdown) , at boot all related interfaces and subif herits mtu (includes bond, bridges, vxlanif, tap...)
On PVE7 (ifupdown2) , at boot bond-if herits mtu but strangely some bridges herits , not other.
It results arbitrary tap and vxlanif have expected MTU
I am investigating this behavior