Open Stack

Problem:

VMs launch, get local IPs from DHCP, get floating ips, and are
available from the external network, but seem to time out after a
minute of inactivity.  After that time, the VM cannot be reached.  I
can sometimes network access back by pinging the external gateway from
the VM console, but eventually times out as well.

Background:

I have Grizzly up and running on Ubuntu 13.04 with gre-tunneling and
per-tenant routers.  I have a network node with three interfaces,
public, management, and vm config.  The compute nodes and quantum node
interact on the vm config network.  Everything else runs from the
controller node.

Here's a diagram of my setup:
http://chavezy.files.wordpress.com/2013/03/ostack-log-net_iscsi.png

I can trace the packet to the external port of the tenant router on
the network node.  So the network node arps for the floating IP with
its interface, no problem.  But when I sniff the tenant network side
of the router, I see unanswered arp requests for the VM's local IP.

What's failing here? How would you troubleshoot this?  Thanks.

--
\*..+.-
--Greg Chavez
+//..;};