Open Stack

Hi Yong,

Garry has recommended that I will you the following:

In: /opt/stack/nova/nova/network/neutronv2/api.py
In the def _get_available_networks function, the developer has added a specific line of code filtering networks by the tenant_id.
Around line 123: search_opts = {"tenant_id": project_id, 'shared': False}

As far as I understand, Neutron already filters non-shared networks by the tenant ID, so why do we need this explicit filter, even more, I think that the behavior of neutron will also return the shared network in addition to the private ones by default so instead of the code doing two calls it could only do one call to Neutron with if needed filtering by net_ids.

Do you see a reason why the code should remain as is?

Thanks,
                -Sam.



From: Samuel Bercovici
Sent: Thursday, August 01, 2013 10:58 AM
To: OpenStack Development Mailing List; sorlando at nicira.com
Subject: Re: [openstack-dev] [Neutron]Connecting a VM from one tenant to a non-shared network in another tenant

There was another patch needed:
In: /opt/stack/nova/nova/network/neutronv2/api.py
In the def _get_available_networks function, the developer has added a specific line of code filtering networks by the tenant_id.
In general as far as I understand, this might be unneeded as quantum will already filter the networks based on the tenant_id in the context while if is_admin, will elevate and return all networks which I belive is the behavior we want.

Do you think this can somehow be solved only on neutron side or must it also be done by rmoving the tenant_id filter in the nova side?

When removing the filter of tenant_id + the pathc bellow, I get the behavior that as admin, I can createVMs connected to another tenants private network but as non-admin I am not able to do so.

Regards,
                -Sam.


From: Samuel Bercovici
Sent: Wednesday, July 31, 2013 7:32 PM
To: OpenStack Development Mailing List; sorlando at nicira.com<mailto:sorlando at nicira.com>
Subject: Re: [openstack-dev] [Neutron]Connecting a VM from one tenant to a non-shared network in another tenant

Hi Slavatore,

I thought that creating a qport  would be enough but it looks like I still missing something else.
I have commented in /opt/stack/quantum/neutron/api/v2/base.py in the create function the ._validate_network_tenant_ownership call.
I can now as an Admin user, can create a qport from tenant-a that is mapped to a private network in tenant-b.

The following still fails with ERROR: The resource could not be found. (HTTP 404) ...
nova boot --flavor 1 --image <image-id> --nic port-id=<port-id>
Where <port-id> is the one I got from the port-create

Any ideas where I should look next?

Regards,
            -Sam.


From: Salvatore Orlando [mailto:sorlando at nicira.com]
Sent: Wednesday, July 31, 2013 5:42 PM
To: OpenStack Development Mailing List
Subject: Re: [openstack-dev] [Neutron]Connecting a VM from one tenant to a non-shared network in another tenant

Hi Sam,

is what you're trying to do tantamount to creating a port on a network whose tenant_id is different from the network's tenant_id?
We have at the moment a fairly strict ownership check - which does not allow even admin users to do this operation.

I do not have a strong opinion against relaxing the check, and allowing admin users to create ports on any network - I don't think this would constitute a potential vulnerability, as in neutron is someone's manages to impersonate an admin user, he/she can make much more damage.

Salvatore

On 31 July 2013 16:11, Samuel Bercovici <SamuelB at radware.com<mailto:SamuelB at radware.com>> wrote:
Hi All,

We are providing load balancing services via virtual machines running under an admin tenant that needs to be connected to VMs attached to a non-shared/private tenant network.
The virtual machine fails to be provisioned connected to the private tenant network event if it is provisioned using the admin user which has admin role on both tenants.
Please advise?

Best Regards,
                -Sam.


_______________________________________________
OpenStack-dev mailing list
OpenStack-dev at lists.openstack.org<mailto:OpenStack-dev at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130807/0da3bf2b/attachment.html>