Hello,

I wasn't sure if this question was best placed in the server or cloud location of the forum.
I am setting up a OpenStack Liberty proof of concept with the Ubuntu 14.04 guide, but i think this problem is on HW/server level.

After much trouble I have connected the two servers direct to each other. In this way the switch was ruled out (VLAN issues) as troublemaker with the ARP trouble. I tested the fysical connections from the compute node eth1 to the network node eth1. The ping command returned "destination host unreachable". So the fysical interfaces wouldn't ping to each other.

If the compute node pings the network node there is a "destination host unreachable" message. At the same time there is a TCPdump listening on the eth1 of the compute node. It receives the ARP-request messages but it won't return them. If the network node is pinged from the compute node (other way around), there is the same issue. The ARP tables on both nodes show incomplete HW address for eth1. So there is something wrong with these nodes or with the network.

compute node = IP 10.0.2.3/24 static (no DHCP)

Network node = IP 10.0.2.2/24 static (no DHCP)

I have checked if there was a IP-address conflict, but there isn't.

The TCPdump output on the network node if the compute node pings:

16:31:57.477751 ARP, Request who-has 10.0.2.6 tell 10.0.2.2, length 46
16:31:57.477766 ARP, Reply 10.0.2.6 is-at 78:e3:b5:fd:72:78 (oui Unknown), length

But the ARP reply never arrives...

Thanks!

Do you have multiple interfaces in the node? Show the outputs of:

Thank you for your reply Smallpond.
I found the solution for the problem. Instead of creating the external bridge on the external network of the network node i created the bridge on the tunnel network. This conflict was the reason why the ARP-resolution was acting so strange.

It had to be like this: (X is the ext-bridge)

Compute node <--> Network node X<--> External network

But it was like this:

Compute node <-->X Network node <--> External network

This thread can be locked down