Strange networking problem

IAmATeaf · 27 Mar 2009 at 15:17

I have 2 subnets, 192.168.1.0 and 192.168.0.0

I have a Windows Server 2003 x64 SP2 server which runs a series of VMs, a NAS, an XP SP2 desktop and a XenServer all on the 192.168.1.0 network. The Windows server also has a second NIC on the 192.168.0.0 network, this is used to route between the 2 subnets.

Problem is that the Windows 2003 server can't talk or ping the XenServer and vice versa but both can ping the NAS and the desktop. VMs running on the Windows server can also ping the XenServer as can PCs on the 192.168.0.0 network via the router. The XenServer can also ping the VMs running on the Windows 2003 server.

I've checked everything including firewall which is disabled, pings just return request timed out from the server and Destination Host Unreachable from the XenServer.

Anybody have any ideas what's going on, pulling out what little hair I have left here.

IAmATeaf · 27 Mar 2009 at 15:38

The XenServers gateway is 192.168.1.200 which is the address of a VM that I use to perform routing, this address is bridged to the actual NIC on the server whose IP is 192.168.1.254. This VM has dual NICs, the other is 192.168.0.200 which again is bridged to the physical NIC 192.168.0.254 on the server. The XenServer can ping both these virtual NICs.

Thing is this all used to work but I had x32 version of Windows Server but the box has 6gb RAM so thought I'd move to x64 so I can utilise all of the RAM.

IAmATeaf · 27 Mar 2009 at 15:50

route print has nothing except the standard paths for both NICS:

Windows Server 2003 x64:

NIC1 192.168.1.254 mask 255.255.255.0
NIC2 192.168.0.254 mask 255.255.255.0 def gateway 192.168.0.1

VM on this server that performs routing, both NICs are bridged to the above NICs:
NIC1 192.168.1.200 mask 255.255.255.0
NIC2 192.168.0.200 mask 255.255.255.0 def gateway 192.168.0.1

XP SP2 Desktop:
NIC1 192.168.1.103 mask 255.255.255.0 def gateway 192.168.1.200

XenServer:
192.168.1.240 mask 255.255.255.0 def gateway 192.168.1.200

As I've said, I've checked and doubled checked everything, the only thing that's changed is the Windows server.

IAmATeaf · 27 Mar 2009 at 15:52

branddaly said:
So if I've got this right, the Xenserver can ping 192.168.1.200 (virtual NIC) , but not 192.168.1.254 (real NIC) ?

Can you confirm the subnet mask and IPs that each of the NIC's has programmed?

Yes that's correct, the XenServer can ping both virtual NICs 192.168.1.200 and 192.168.0.200 but not the real 192.168.1.254 and 192.168.0.254, it's bloody driving me nuts.

IAmATeaf · 28 Mar 2009 at 22:01

Well I installed WireShark on the 2003 server to see what's happening, when I ping the XenServer WireShark reports ARP requests asking who knows what 192.168.1.240 (the XenServer) is so no ICMP packets seem to be sent. When I try the ping from the XenServer I seem to get nothing but this could be because I'm monitoring a particular NIC on the Windows Server and both ARP and ICMP are broadcasts?? Maybe?

As you can tell I'm not an expert when it comes to network traces.

IAmATeaf · 30 Mar 2009 at 19:30

I sort of figured out what's going on but have no idea how to fix it. The onboard Marvell NICs are what's causing the problem. Accidentally discovered it when I decided to use my home PC to manage the Xenserver as it has a 24" screen instead of the Windows 2003 server. The home PC couldn't ping the Xenserver either, the common factor being that the Windows 2003 server has an onboard Marvell 88E8055 NIC, the Xenserver has an Asus P5Q Deluxe mobo as does my home PC which have the Marvel 88E8086 NIC.

If you play around with driver settings on the server/home pc eventually it'll ping the Xenserver but then at some random time it'll stop.

Installing an Intel based 100mb NIC resolved the problem, so installed 1 in the Xenserver and in the Windows 2003 server and everything talks to everything else. Shame really as I was hoping to make use of the onboard 1gb.