Confusing Network Interface issue

[mccont]

Hi,

I have a one of those annoying network connectivity issues that I'm having trouble understanding.

For the moment, let's deal with the machine exhibiting the problem.

It is a Mac Mini, running OS X 10.8.4. I'm running VirtualBox 4.2.16 r86992 with the corresponding extension pack.

The host machine is configure for both Ethernet and Wireless networking, with IP addresses 192.168.0.240 and 241 respectively. There are no firewalls running, there are no other hosts using these IP addresses.

I can ping/ssh/whatever from every other physical host in the network without any problems.

On this host I have a number of Ubuntu guest OS. They are all cloned from the same original configuration, using Bridged networking on the ethernet port, all with unique MAC addresses.

Since I'm testing a cluster product, these are configured in groups, typically of three hosts, with fixed sequential IPs. For example, I'll have host ct1, ct2, ct3 with fixed IPs of 192.168.0.200, 201, 202. They can all ping/ssh/any network traffic each other without issue. They can also also access the router (192.168.0.1), access the internet, and indeed any other host on the network without any issues.

Note, there are *NO* IP configuration on these machines - they ping all the hosts active through the subnet without any issues, including addresses above and below the host IP address.

The problem manifests itself in two ways:
+ The guest OS *cannot* see the host machine on the ethernet port. Pinging, ssh, etc to 192.168.0.240 reports 'network host unreachable' or simply gets 100% packet loss. All other IPs work fine.
+ Intermittently, I cannot ping any guest VM on this host from other machines in the network, but this can be variable

To further add to the confusion, an identically configured Mac Mini, on the same network switch, with the same virtual machines (either cloned by copying, by doing an export and import) exhibits none of these problems. All the VMs on this host can ping the VM host and all other hosts in the network. Another host, also on Mac OS X (iMac) has 50 VMs based on this same cloned guest VM, none of them exhibit this problem either.


Things I've tried:

+ Reconfiguring the host IP address, MAC address and wireless configurations of the host OS
+ Reconfiguring the host IP address, MAC address and wireless configurations of the guest OS
+ Hose networking (with port redirection) works fine
+ Using the Wireless (en1) as the bridged adaptor in place of en0 for the VM; this works, but leads to exceedingly slow networking (not good for network performance testing)
+ Forcibly flushing and populating the ARP tables. Interestingly, the ARP output correctly shows the host ethernet IP and MAC address without issues
+ Changing the network cables, switch, and switch ports
+ Restarting every single device in the network, including switching off everything and then powering up individual machines

I'm now out of ideas - especially since it is only this one machine that is affected. If the machine had any physical defects in the hardware, I don't see how I would be able to ping/file copy etc stuff to it, so I'm loathe to suspect a hardware failure. And obviously there's no network failure for the VMs themselves as they are able to see everything else on the network.

Any ideas gratefully received.

MC

[BillG]

I don't see how you can rule out a hardware problem when the hardware seems to be the only thing which is different over the three systems. Problems of this nature are usually network card or network card driver issues rather than software or configuration problems.

Communication between the host and machines on the virtual network depend on the VirtualBox bridge driver working with the NIC driver in the host OS to enable the vms to use the physical NIC. It is essentially a filtering process to ensure that the packets get to the correct MAC address. I have seen problems like this fixed by updating the NIC driver in the host OS.

Are all the MAC Minis the same type and running the same OS version?

[mccont]

Yes, two identical Mac Minis (CPU, HD, RAM) and identical OS X.

The problem with accusing the hardware is that I would expect it to manifest itself somehow in connectivity problems to the host and/or problems with the guest OS communicating out to other hosts in the network (which works fine, just connecting in is faulty). In 25 years of network admin, IME, network cards don't get selective like that.

I've killed and recreated the ethernet configuration within OS X, and I'm loathe to do a complete OS reinstall at this stage just due to time.