VLAN for management interface breaks networking if vm's use the same VLAN

[troykelly]

I'm hoping somebody has a fix for this, I created a bug report with the details here.

If the Proxmox VE 8.x machine is managed by a VLAN, let's say `100` for example, and then a virtual machine uses that same VM all networking on the host machine dies.

It's super easy to reproduce if anybody wants to test it in a lab, but I can't work out what the fix for it is.

No, I can't set the management network to be untagged for the proxmox ve hosts, management has to be over a VLAN.

 

[Yogorov]

do you have switch with vlan function?
do you plug pve8 at switch trunk port?
can you show your /etc/network/interfaces content?

 

[troykelly]

The technical details are in the bug report that I linked to previously.

I'm not sure what you are asking, so I will re-write the question and answer it - if I have misunderstood, please let me know.

Does our network / switching infrastructure support VLAN's?
Yes, they are used extensively.

Is the Proxmox VE server connected to switch port with trunking enabled?
Yes.

> can you show your /etc/network/interfaces content?
Yes, it's detailed in bug report 5043.

To be clear - VLAN's are working just fine on the Proxmox VE host (and the rest of our network)
The issue is that a virtual machine with a VLAN of the same number as the management VLAN causes all networking on the host to fail.

 

[Yogorov]

here is my interfaces for your reference

auto lo
iface lo inet loopback


#Intel X540-AT2
iface enp65s0f0 inet manual
iface enp65s0f1 inet manual



auto bond0
iface bond0 inet manual
    bond-slaves enp65s0f0 enp65s0f1
    bond-miimon 100
    bond-mode 802.3ad
    bond-xmit-hash-policy layer2+3
    hwaddress a0:36:9f:7e:f4:18

auto vmbr0
iface vmbr0 inet manual
        bridge-ports bond0
        bridge-stp off
        bridge-fd 0
    hwaddress a0:36:9f:7e:f4:18

auto vmbr0v12
iface vmbr0v12 inet static
        address  192.168.12.55/24
        gateway  192.168.12.1
        bridge-ports bond0.12
        bridge-stp off
        bridge-fd 0
    hwaddress a0:36:9f:7e:f4:18

my pve7 at switch trunk port and it has ip at vlan12
at switch side I setup LACP for this pve7

all my vm has vlan tag 12 work fine

I saw your interfaces has two port at one bridge interface
maybe you can try LACP with 802.3ad and with switch setting

 

[troykelly]

Do you think your double bridge (bond0 --> vmbr0) is what is preventing the networking failure that I am seeing?

We don't use LACP, the bridge is in case of a catastrophic failure of our 10Gb infrastructure, allowing us to fallback to 1GbE to keep systems online.

Is there a particular reason you have a bridge in a bridge configuration rather than just one bridge? I'm struggling to find any best practices for this, and there's not a lot of documentation at all - so I'd like to know if we are doing this completely wrong

 

[Yogorov]

yeah, I use LACP for two things
1. 10G+10G so I got 20G bandwidth
2. prevent network failure , like: unplug one wire , server still online