Configure compute node in OpenStack Installation Guide for Ubuntu 12.04/14.04 (LTS) - icehouse: share_dhcp_address should not be True

[Dražen Lučanin]

By following the advice from the docs and setting share_dhcp_address = True I got the same IPs for my controller and compute nodes. This resulted in getting messages "no route to host" when trying to ssh from the controller node to a VM running on the compute node. This should be possible and therefore the share_dhcp_address should be set to False (before creating the network).
-----------------------------------
Built: 2014-09-19T07:50:46 00:00
git SHA: 7241eb0fbd6dedac9007a28863fe7dcbf361e96f
URL: http://docs.openstack.org/icehouse/install-guide/install/apt/content/nova-networking-compute-node.html
source File: file:/home/jenkins/workspace/openstack-manuals-tox-doc-publishdocs/doc/install-guide/section_nova-networking-compute-node.xml
xml:id: nova-networking-compute-node

[Dražen Lučanin (kermit666)]

When I changed this setting, deleted the old network (using nova net-delete), recreated it and restarted both machines I managed to get controller-vm@compute connectivity.

[Matt Kassawara (ionosphere80)]

Upon further investigation, your observation seems like an intended side-effect of the 'share_dhcp_address' option. Can you address the following items?

1) Provide a general description of the network interfaces and networks on each host.
2) Provide the contents of nova.conf on the compute node.
3) Provide the 'nova network-create' command used to create the tenant network.
4) Provide the 'nova boot' command used to launch the instance.
5) Did you assign a floating IP address to the instance?
6) What source/destination IPs did you use to reach the instance?

[Dražen Lučanin (kermit666)]

I was following the OpenStack Icehouse installation guide on Ubuntu 14.04 to the letter (the legacy networking route) - the same 2 physical machine configuration that is recommended in the guide and the same network creation and booting commands listed there.

http://docs.openstack.org/icehouse/install-guide/install/apt/content/

I did not set up floating IP addresses as that is not explained at that point of the documentation yet.

The documentation page where it says I should be able to ping or ssh the VM "from the controller node or any host on the external network" shows that what I am experiencing by following the documentation (only being able to reach the VM from the node that is physically hosting it) is not expected behaviour.

http://docs.openstack.org/icehouse/install-guide/install/apt/content/launch-instance-nova.html

I used the IP address of the VM shown in the "nova show" command. Like I said, when I flipped the flag back and recreated the network, everything worked, so I think this is a documentation bug - the flag you are currently suggesting should not be set.