VirtualBox 2.2.4 and bonded host interface with guest bridge networking

[John Huong]

My host has two Gigabit NICs. It runs on 64 Bit Ubuntu 8.0.4.2 Server. I have bonded the two interface to become bond0.

I created another guest using 32 Bit Ubuntu 8.0.4.2 Server and configured it to be bridged with bond0.

It seems to work for a while, but halfway through the guest networking stops responding (could not ping to the network gateway or access the Internet from the guest). The host network works well as I was accessing the guest through remote desktop.

[John Huong]

Also the bonding method I have chosen is balance-alb.

When it is in balance-alb mode, the problem is present.

Changing it to mode=1 will make the problem go away, but I no longer have the benefit of load balancing across two NICs.

[Technologov]

This bug is remotely related to bug #5519

Flancer: Does it happen with both NAT networking and Bridge networking ?

-Technologov

[John Huong]

I have only tested with Bridge networking. I no longer have access to those systems.

[Technologov]

Nם Hardware = No Test. I suggest closing this bug.

-Technologov

[dbmaxpayne]

Hi,

same problem here.

I'm using Ubuntu 10.04 and also created a bond0-interface. First I wanted to use bond_mode 6 (balance-alb) for load-balancing but unfortunately the guest-OS (Windows XP) stops responding after a few seconds. If I de- and reactivate the guest-NIC it responds again for a few seconds.

After setting bond_mode to 0 (balance-rr) the guest-OS responds better. But after some hours it also looses connectivity.

Virtualbox Version is 3.2.4. Guest-Network is bridged to interface bond0. GuestAdditions are installed.

Mark

[Hervé Pellan]

Hi,

I also posted my message on: ​http://www.virtualbox.org/ticket/5260#comment:24[[BR]] Because the syslog of the host mentioned clearly an allocmem error of VBoxNetFlt.

But... I'm still guessing it could be related to the way network interfaces are binded between host and guest.

I noticed that, when the guest OS starts, it does something (wrong?) to the bonding, which one of the network card goes into promiscuous mode. As if the virtual card tries absolutely to bind to a physical card.

The bonding is still functionnal since both host and guest are responding to network requests. As these systems are in production, it's delicate to verify if the bonding is still really working (maybe it is working only on one physical card after the other goes into promiscuous mode).

I first thought that the vboxNetFlt allocmem error occurs because of the particular case of bridged over bonding, but I have another system (pretty much identical) with same network config which doesn't cause any allocmem error, but has 4GB real RAM.

Host:
-- Linux file server (on which the vboxNetFlt allocmem error occurs), DNS secondary.
-- System is OpenSuSE 11.2 x86_64 kernel 2.6.31-12.
-- Vbox 3.2.10-109.3.x86_64 running in headless mode.
-- Doesn't have much real memory: 2GB, backuped with a 6GB swap space.
-- Main network interface is a bonding of two Broadcom BCM5780 Gigabit.
-- This host doesn't do much, but as a backup server, it has an intensive activity at night when rsync script starts. Mostly, this is at this time that the vboxNetFlt allocmem error occurs.

Guest:
-- System is Ubuntu 9.10 x86_64 kernel 2.6.31-20.
-- Network card is bridged on this bonding with the virtual Intel 1000e driver.
-- The guest OS, as opposite, has a continuous intensive network activity: access to shared files, file sharing itself and master network service for computing dispatcher.

Thanks for any clue.

[Sean Pappalardo]

I have this same problem on Debian Squeeze AMD64 with VirtualBox v4.1.0. Tickets #8866 and #5747 may be duplicates of this one.

When I switch to NAT, everything works fine. It's only in bridged mode that the problem occurs.

[Sean Pappalardo]

Log of VM session on VB v4.1.0 on Debian Squeeze AMD64 wherein the problem occurred.

[Sean Pappalardo]

I also just tried setting Promiscuous Mode to "Allow All" but it didn't seem to help much.

[Sean Pappalardo]

Workaround: changing the bonding mode on the host to balance-tlb (mode 5) seems to work fine for me with VB 4.1.0. This bug still needs to be fixed though since this mode is not as fault-tolerant as balance-alb.

[scurvy]

This problem is pretty easy to fix, but it is on the VirtualBox side.

The problem is that virtualbox only goes into promiscuous mode on the first device in a bond setup. If you have 2 or more active adapters, it won't turn on promiscuous for the other adapters.

If VirtualBox would enable promiscuous mode for all adapters in the bond, bonding with balance-alb would work.