OpenVZ Unable to configure IP addresses on 2 different subnets

[dgardner]

I have a system running CentOS 6.2 and OpenVZ 3.0.30. The system has 2 NICs on 2 different subnets as follows:

Eth0: 192.168.5.30/24 (GW is on this subnet)
Eth1: 192.168.6.30/24
venet0

I can build a container and give it an IP address on the 192.168.5. subnet and it works fine.

When I try to add another IP on the 192.168.6. subnet it fails with:

Adding IP address(es): 192.168.6.51
arpsend: 192.168.6.51 is detected on another computer : 00:30:48:67:3c:e4
vps-net_add WARNING: arpsend -c 1 -w 1 -D -e 192.168.6.51 eth0 FAILED

The problem is that it tries to add it on Eth0 in the 192.168.5. subnet instead of Eth1.
The duplicate detection error is bogus as this IP does not exist anywhere else.

I've read through a lot of items but can't quite nail down how to resolve this scenario.

I need this container to talk on both subnets if possible - One is Orange (to the firewall and out to the Internet), and the other is Green (no internet, backbone between SQL, shared drives, etc.)

Any help is appreciated.

[falko]

Do you have

NEIGHBOUR_DEVS=all

in /etc/vz/vz.conf?

[dgardner]

Yes, "NEIGHBOUR_DEVS=all" was set in the /etc/vz/vz.conf file and the server has been restarted since it was set so I know it's in.

I tried running arpsend myself and got the following results:

arpsend -D -e 192.168.6.50 eth0 (Note: 192.168.5.x/24 interface)
arpsend: 192.168.6.50 is detected on another computer : 00:30:48:67:3c:e4

arpsend -D -e 192.168.6.50 eth1 (Note: 192.168.6.x/24 interface)
(This request hangs indefinitely)

I do not know who has a Mac of "00:30:48:67:3c:e4". I have not been able to locate it on any of my other systems and I can't believe there would be any 192.168.6.x systems on the 192.168.5.x network. It would almost have to be running in promiscuous mode as any IP I try arpsend with returns in use.

So it looks like arpsend is returning a detected message on the first interface and that it does not proceed to the second. Never the less there are two issues as I see it:
1. Why is arpsend returning a conflict on eth0?
2. Why does arpsend hang when I try it on eth1?

So I am guessing that both of these will need to be resolved. Hopefully this information will spur some thoughts from others. Thanks for your reply.

[dgardner]

More info: I actually put in two identical servers a week ago, both CentOS and OpenVZ. Turns out the other server (which I haven't been using) had the mysterious MAC address of "00:30:48:67:3c:e4" on eth0. I brought that interface down and now OpenVZ on the first box initializes fine with both IP addresses. The IP config on Box #2 looks normal and the Eth0 interface is defined as 192.168.5.41/24. So why is it responding to an ARP request for 192.168.6.50? Seems really odd. So now OpenVZ looks great but I have to figure out how to bring up the other box. Odd...