Internal networks stop passing traffic

[Matt Kassawara]

I'm attempting to simulate a dynamically routed network environment by running Quagga on several minimal Ubuntu 9.10 64-bit Linux guests hosted on MacOS X 10.5.8. Each guest uses a handful of internal networks to simulate point-to-point links between routers and to provide connectivity for clients which generate traffic on stub networks. Shortly after the guests boot up and routing converges, the internal networks either become incredibly sluggish or completely fail to pass traffic. Running "ping" between guests shows considerable packet loss and "tcpdump" on each side shows packets egressing but few, if any, returning from the remote end. Sometimes I'll see "bursts" of packets, but in general very few successfully complete the round trip in a reasonable time. Strangely, keeping the internal network links busy after the guests boot up with something like a 1-second interval ping maintains connectivity. However, the internal networks will stop passing traffic shortly after terminating the ping. None of the guest log files show any problems. I'm also experiencing this issue with guests not associated with this particular project, but running under the same instance of VirtualBox.

[Andreas Böckler]

I have the same issue with 3.1.4

host only networking stops working without a error: I've mounted my /Users from HOST--OS X 10.6.2 on a Debian32bit-VM via NFS.

Suddenly the network locks up. NO PING .. nothing is working

VBoxManage controlvm ... setlinkstate2 on/off has no effect (but is seen from the virtual machine)

[Frank Mehnert]

Please attach a VBox.log file of such a session. How easy to reproduce is the problem and how long does it take until the network does not respond anymore?

[Andreas Böckler]

Hi,

it seems to be a problem with e1000-Emulation. I've removed the e1000 driver from the linux/kernel and loaded it again. Problem was still there.

I've switched to virtio and the problem is gone.

[Andreas Böckler]

VBox.log with broken host-only adapter

[Andreas Böckler]

the problem still arise: Setup: debian-vm on osx 10.6.2 eth0 -> bridged en1 (airport) - type virtio-net eth1 -> hostonly - type virtio-net eth2 -> bridged en0 (ethernet) - typo virtio-net

eth1 stops working after some time (can be 10 minutes or more)

• /sbin/reboot of guest doesn't help for interface eth1 guest sees following at reboot

DHCPDISCOVER on eth1 DHCPOFFER FROM x.x.x.x DHCPREQUEST on eth1 <repeating forever>

[Frank Mehnert]

Your suspicion of the E1000 device still holds, correct? We have other reports that switching to PCNet solves this problem as well.

[Andreas Böckler]

I don't know ... all 3 devices are virtio-net now and it happens too.

Now it seems stable since for 8 hours, but the i didn't switch on the eth2-interface (en0-ethernet on host)

An observation: The ethernet-Interface on the macmini is connected (without a switch) to a sat-Set-Top-Box, which is not always on. When the set-top-box is switched off, the cable is not connected for the macmini -> bridged networking stops on eth2. But this shouldn't interfere with eth1 (which is the hostonly network), should it?

[Andreas Böckler]

output of "VBoxManage showvminfo darkside --details"

[henrik242]

This looks like a dupe of #5254

[Frank Mehnert]

Still relevant with VBox 4.0.4?

[Frank Mehnert]

No response, closing.