Cannot ping instance after creating vm (using quantum networking)

1.when u have two networks, you should specify which network to boot a vm.
nova boot --image xxx --flavor 1 --nic net-id=<net-id> myserver1
by default, nova allocates one ip for each of available networks. So it is not about auto_assign_floating_ip
2. to fix it, we have to know what configuration in nova.conf, quantum.conf and L2 configuration

I think step #3 was a mistake.

3) quantum net-create ext_net --shared --router:external=True

Should be:

3) quantum net-create ext_net --router:external=True

I.e., the network should be be shared.

Shared allows VMs to plug directly in, whereas we want ext_net to purely be external (but not shared).

When a VM is booted and no --nic is provided, it will be connected to all networks it sees, namely those that are owned by the tenant booting the VM, as well as all shared networks. This is why your VM has a NIC on ext_net when it shouldn't. The VM has no floating IPs.

Also, the reason you likely can't reach your VM via SSH on the internal IP may be security groups.

Thanks for the tips so far! After setting the --nic option on "nova boot" and then re-issuing the "quantum net-create" without the --shared option, I was able to progress farther down the instructions, including issuing floating ips. While I can now ping, it would seem that I cannot ssh. I tried manually adding the floating ips to the br-ex bridge, but that did not seem to help. Here are some observations:

* From the quantum network node and quantum server (on the nova controller), I can ping any ip between 172.16.1.1 - 172.16.1.101.
* From the compute node, I can only ping 172.16.1.1. Any other 172.16.1.x ip results in a "sendmsg: Operation not permitted"
* From the quantum network node, I can ping 1.1.1.131 and 1.1.1.132; however, I cannot ping those ips from the quantum server or the nova compute node.
* I cannot ssh into any instance, though I added tcp/22 into the default security group (using nova, not quantum)

I will upload the config files shortly.

Here are the outputs from some commands:

http://paste.openstack.org/show/32572/

Other notes:
* I created two vms
* On the quantum network node, I attempted to add the floating ips using 'ip addr', e.g. "ip addr add 1.1.1.131/24 dev br-ex". The ips added to br-ex, but I could not ssh into the vms
* I attempted to use "route add -net 172.16.1.0/24 gw 172.16.1.1", but I received "SIOCADDRT: No such process". It didn't add, but do I need to do this?

Again, thank you for the help,
Ed

http://paste.openstack.org/show/32695/

This is a paste of the following config files:

nova controller's nova.conf
quantum server's quantum.conf (on nova controller)
quantum server's ovs_quantum_plugin.ini (on nova controller)

network node's dhcp_agent.ini
network node's l3_agent.ini

network and compute nodes' quantum.conf
network and compute nodes' ovs_quantum_plugin.ini

Are there any other configs that would be useful to provide?

The problem still exists..

1.
> From the quantum network node and quantum server (on the nova controller), I can ping any ip between 172.16.1.1 - 172.16.1.101

What does mean you can ping any of those ips? Do you mean that you don't get "Operation not permitted"?

2.
> From the compute node, I can only ping 172.16.1.1. Any other 172.16.1.x ip results in a "sendmsg: Operation not permitted"

You may try the following commands:
1) ip netns
 will give you a list of network namespaces.
Choose one that is related to tenant subnet (172.16.1.x) - it would be either qdhcp-<network_id> or qrouter-<router_id>
where <router_id> is a router attached to 172.16.1.x network

2) sudo ip netns exec namespace ssh 172.16.1.2
Where namespace is eiath qdhcp-... or qrouter-.. from previous command

3.
> From the quantum network node, I can ping 1.1.1.131 and 1.1.1.132; however, I cannot ping those ips from the quantum server or the nova compute node.

Sure you can ping them since it's ips on one of the local nics (br-ex). No other hosts can see them since they don't share 1.1.1.0/24 network with quantum network node.

1) Seems that your br-ex has 10.140.65.63 which is a real external address.
2) I'm wondering why you specify subnet for external network as 1.1.1.0/24 | {"start": "1.1.1.130", "end": "1.1.1.249"}
It looks like it should be something like 10.140.65.0/??

I have since switched my configuration to a shared flat network in order to try a simpler setup. I will revisit this question if I can get the simple case working. Thanks!

Ultimately, the cause for this problem was that the dhcp agent kept on dying immediately after starting up. It was not immediately apparent since no errors were made in any logs. After fixing a file permission on the filesystem in /var/lib, the dhcp agent ran fine and the instances received the ips as expected.

any chance you could post the exact command which fixed this, i was looking at the bug and wanted to see if i could reproduce it.

chown -R quantum.quantum /var/lib/quantum

I installed from the folsom packages from the ubuntu cloud archive, using 12.04.

ok good, thanks, will see if i can reproduce it locally here.