Cannot attach CentOS VMs via DHCP following router swap

Question

I recently swapped out an old Netgear modem/wireless router and replaced with a ZyXel P660HW-T1 supplied by our ISP. I have configured it so that the new router has the same IP address, subnet mask, IP Pool starting address and IP Pool size as the old one. In addition I configured it to use the same wireless SSID and key as the old device. (To stop the users shouting at me:-))

Windows machines connect to the network (all over WiFi) without difficulty. Two Ubuntu machines connect via WiFi without difficulty (the IP address is assigned to wlan0, which I guess is what one would expect).

The problem lies with CentOS VMs which we run on the Windows workstations. Previously we could connect these in Bridged mode and they would be assigned an IP address on the network by DHCP. The device attached was eth0. Now however, this works only very occasionally. Given that I think I have configured the new router identically to the old one, I can't understand what the problem is. We don't want to roll back to the old router as it had got unreliable, regularly dropping connections and so on. Anybody got any suggestions?

Answer 1

Turns out the problem was with the ZyXel router itself in this case - it seems the initial batch of this particular model was sent out with a dodgy firmware which couldn't handle DHCP leases properly. After about two weeks of phone calls(!) and frustrated users I managed to get a replacement sent out from the ISP and it worked perfectly out of the box...

Answer 2

For debugging network problems tcpdump (Linux) and Wireshark (Linux and Windows) are your friends.

Wireshark is full of security vulnerabilities (typically buffer overflows in some of its hundreds decoders). I'm not particularly worried about them, but in order to be on the safe side I prefer to do the capturing with the dumber

sudo tcpdump -s 0 -i eth0 -w capture1.cap

and use the luxury of wireshark on the captured file just as an ordinary user.

I haven't used Windows for a couple of years, I'm not sure whether you will see bridged traffic on the windows host. If you don't see it you can capture on the guest while you bring networking up. In cases where capturing does not succeed before the network is up I have written an endless loop to try to start tcpdump until it succeeds (i.e the network has come up).

In CentOS networking is brought down and up using the commands

sudo service network stop

sudo service network start

Probably you want to look at the DHCP traffic.

Of course reading Wireshark output requires basic understanding of how the underlying protocol (e.g. DHCP) works. The net is full of explanations. Also comparing the traffic from/to a machine that has no problems (or your old router during a quiet hour) can help to understand.

Happy debugging!