0

I have two identical VMs (Windows XP guests). I use KVM for the hypervisor. The first one is running, the second is a clone of the first not yet running. I edited the config for the second one to put network link state as down and give it a unique MAC address. The first VM has a static IP set, which means the second/clone VM does too. I powered on the cloned VM with a now unique MAC address, and the link state down, and gave it a unique hostname/netbios name.

I meant to change the static IP to something unique on the second clone but I forgot to, and enabled the link with a duplicate IP address on both VMs. Of course we have dynamic DNS so the second VM overwrote the DNS host record, now with its own name. That was expected. What was unexpected, is that after changing the second machine's ip to be unique, I still could not ping the first machine by IP. Why? My guess is layer 2 caching on the switch, where the switch had the MAC address of VM number 2 being tied to the IP address that is now the IP address of machine 1, and even though I changed the IP on VM2, the switch wasnt aware of this.

  • Are both VM's on the same host? Do they connect to the same physical switchport? This doesn't feel like a switching issue to me. Can both machines ping their gateway? – Dan Apr 26 '13 at 21:59
  • Different Hosts. I dont know about which switch they were connected to. –  Apr 26 '13 at 23:01

2 Answers2

1

Not the MAC table on the switch, as that was always correct. What may have happened is that the ARP cache was not updated?

Jim G.
  • 2,607
  • 1
  • 18
  • 19
0

This can be because of Windows. At least Windows Server were disabling network access when detecting that there's an IP conflict during boot, so if you rebooted the first one while having the IP conflict it never got networking enabled.

V13
  • 231
  • 1
  • 5
  • The first machine, that was already booted and online is the one that I lost access to after booting the second one, which I never lost access to. –  Apr 26 '13 at 23:00