Controller isn't forwarding packets

[gonçalo Semedo]

Hi,

I created a slice using flack named miniWan in the project LoadBalancingRealSDNs_FCUL.

The slice have only VMs from instageni.stanford, instageni.illinois and instageni.clemson.

I used egre-tunnels to connect each vm because stitching wasn't working, by the way what is the difference between egre-tunels, gre-tunnels and stitch?

The vm with the floodlight controller has an public ip.

Each ovs switch successfully connected to the controller, but i can't ping the hosts.

Can anyone help me?

Thanks
Gonçalo Semedo

[Sarah Edwards]

Hi Gonçalo,

In general, we have found that gre and OVS don't work together.  If you (or anyone else on this list) finds differently, please let us know.  See for example this earlier thread on this topic: https://groups.google.com/d/msg/geni-users/dqHG6I9gGsU/E-1xI4jqyk8J

You asked about the difference between stitching, gre, and egre...

Stitching gives you a layer 2 dataplane connection (in particular it's an inter-domain VLAN).  A list of which sites support stitching is in [1]. Note that Clemson is not on the list of supported stitching sites in [1].  An example of how to do stitching is in [3].

gre-tunnel [2] give you a Layer 3 tunnel running over the control interfaces on your nodes.  Because it's over your control interface you have less bandwidth available.  This is only used to connect InstaGENI OpenVZ nodes.

egre-tunnel [4] tunnels Layer 2 over Layer 3.  It runs over the control interfaces on your nodes.  Again because it's over your control interface you have less bandwidth available.  This is only used to connect InstaGENI xen nodes.

Cheers,

Sarah Edwards, GPO

[1] Stitching sites:  http://groups.geni.net/geni/wiki/GeniNetworkStitchingSites

[2] GRE:  http://en.wikipedia.org/wiki/Generic_Routing_Encapsulation

[3] Stitching example: http://groups.geni.net/geni/wiki/GENIExperimenter/Tutorials/StitchingTutorial

[4] http://www.cisco.com/c/en/us/td/docs/ios-xml/ios/interface/configuration/xe-3s/ir-xe-3s-book/ir-eogre.html 

--
GENI Users is a community supported mailing list, so please help by responding to questions you know the answer to.
 
If this is your first time posting a question to this list, please review http://groups.geni.net/geni/wiki/GENIExperimenter/CommunityMailingList
---
You received this message because you are subscribed to the Google Groups "GENI Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geni-users+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

*******************************************************************************

Sarah Edwards

GENI Project Office

BBN Technologies

Cambridge, MA

phone:    (617) 873-2329

email:    sedw...@bbn.com

[Sarah Edwards]

Hi again,

I spoke with our OpenFlow expert and he says that OVS and egre tunnels (but not the gre tunnels) should in general play well together.  That said, I am unable to get this to work and am suspicious that there is another factor at play.

If I figure out more, I'll let you know.  

In the meantime, depending on your topology I suspect that stitching is the way to go.  

If you send us an RSpec of your topology I can use it for testing the egre/OVS issue and can also give you some pointers about how to best use stitching with your topology.

Thanks,

Sarah

[Leigh Stoller]

> I spoke with our OpenFlow expert and he says that OVS and egre tunnels
> (but not the gre tunnels) should in general play well together. That
> said, I am unable to get this to work and am suspicious that there is
> another factor at play.

They should since the egre tunnels on Instageni XEN nodes are implemented
with openvswitch ...

Leigh

[gonçalo Semedo]

Hi all,

Thanks for your replies I will check that stitching tutorial.

I will send you now my rspec.

I notice that the port 6633 is always in use, so i modified my floodlight to be listening on port 6644. But how can I change the floodlight interface? Because when I use it it shows a different topology than mine.

Thanks again,

Gonçalo Semedo

[gonçalo Semedo]

Hi again,


I changed my topology and now i only use egre-tunnels (I replaced the LANs for egre).

Now the packages are going through the controller and the topology on the web interface is now correct (before no links had appeared).

When I ping, floodlight detects the hosts but I still get Destination Host Unreachable.

I notice that only ARP requests are sent, and no ARP replies.

Does this information help?

Thanks

Gonçalo Semedo

On Saturday, 19 April 2014 16:51:26 UTC+1, gonçalo Semedo wrote:

[Ezra Kissel]

Hi Goncalo,

Are you using the Forwarding module in Floodlight? If I remember
correctly, its default behavior should flood ARP (eth broadcast
addresses) and allow your nodes to resolve each other. If not, you
should see unhandled packet-INs if you run Floodlight at higher debug
levels.

- ezra

On 4/22/2014 1:10 PM, gonçalo Semedo wrote:
> Hi again,
>
>
> I changed my topology and now i only use egre-tunnels (I replaced the
> LANs for egre).
>

> Now the packages are going througth the controller and topology on the
> web interface is now correct (before no links appeared).
>
> When I ping, floodlight detects the hosts but I steel get Destination

> Host Unreachable.
>
> I notice that only ARP requests are sent, and no ARP replies.
>
> Does this information help?
>
> Thanks
> Gonçalo Semedo
>
>
>
>
> On Saturday, 19 April 2014 16:51:26 UTC+1, gonçalo Semedo wrote:
>

> Hi,
>
> I created a slice using flack named miniWan in the project
> LoadBalancingRealSDNs_FCUL

> <https://portal.geni.net/secure/project.php?project_id=4c767671-04c6-418a-bf89-b9a630ba1dac>.

>
> The slice have only VMs from instageni.stanford, instageni.illinois
> and instageni.clemson.
>
> I used egre-tunnels to connect each vm because stitching wasn't
> working, by the way what is the difference between egre-tunels,
> gre-tunnels and stitch?
>
> The vm with the floodlight controller has an public ip.
>
> Each ovs switch successfully connected to the controller, but i
> can't ping the hosts.
>
> Can anyone help me?
>
> Thanks
> Gonçalo Semedo
>

> --
> GENI Users is a community supported mailing list, so please help by
> responding to questions you know the answer to.
>
> If this is your first time posting a question to this list, please
> review
> http://groups.geni.net/geni/wiki/GENIExperimenter/CommunityMailingList
> ---
> You received this message because you are subscribed to the Google
> Groups "GENI Users" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to geni-users+...@googlegroups.com

> <mailto:geni-users+...@googlegroups.com>.

[Sarah Edwards]

Hi Goncalo,

Could you clarify which nodes are your OVS switches?  Is it the three in the triangle (VM, VM-0, and VM-2)?

Thanks,
Sarah

[gonçalo Semedo]

Hi,

Yes Sarah, the 3 vms in the triangle are OVS switches.

Ezkissel, yes I am using the forwarding module.

Thanks

Gonçalo Semedo



On Saturday, 19 April 2014 16:51:26 UTC+1, gonçalo Semedo wrote:

[Sarah Edwards]

Thanks Gonçalo,  That's very helpful.

Could you run a simple test for us?

1) Could you disconnect your nodes from your controller and allow the OVS switches to revert to a standard learning switch?

2) Then see if ping works? It should.  If not, then something more fundamental is wrong.

Thanks,

Sarah 

--
GENI Users is a community supported mailing list, so please help by responding to questions you know the answer to.
 
If this is your first time posting a question to this list, please review http://groups.geni.net/geni/wiki/GENIExperimenter/CommunityMailingList
---
You received this message because you are subscribed to the Google Groups "GENI Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to geni-users+...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[gonçalo Semedo]

Ok I think I found the problem.

The hosts were on different subnets, using wireshark i found that, when I ping, in the wire, the ip had changed to a different one, but in the same subnet.

So i changed the hosts ips, to be in the same subnet and now it seems that is working fine.

Thanks for your help

Gonçalo Semedo
On Saturday, 19 April 2014 16:51:26 UTC+1, gonçalo Semedo wrote:

[Sarah Edwards]

Ah!  That makes sense and it's a good lesson for posterity...  

It is always a good idea to set the IPs in a multi-site topology.  Each aggregate doesn't know what IPs other aggregates are assigning, so letting the aggregates pick in the multi-site case is unlikely to result in a good outcome.

Glad you figured it out!  Let us know if you have any more problems.

Cheers,

Sarah

--
GENI Users is a community supported mailing list, so please help by responding to questions you know the answer to.
 
If this is your first time posting a question to this list, please review http://groups.geni.net/geni/wiki/GENIExperimenter/CommunityMailingList
---
You received this message because you are subscribed to the Google Groups "GENI Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geni-users+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.