IG Utah Xen to IG GPO Xen with GRE tunnel link fails

[lnevers@bbn.com]

Writing ticket to track resolution of issue in email:

Running test case IG-XN-11 ( IG Utah Xen to IG GPO Xen with one GRE tunnel link) and have run into a problem. The two Xen hosts come up without a problem, but when I login to the each of hosts there is no dataplane GRE tunnel configured.

Slice is IG-XN-11 and hosts are:

• pc5.instageni.gpolab.bbn.com port 32314
• pc3.utah.geniracks.net port port 32058

Waited up to 10 minutes in case the configuration portion was delayed, but no change.

Also if the same RSpec is used with emulab-openvz in place of emulab-xen, GRE works.

[lnevers@bbn.com]

Capturing email exchange that has taken place so far:

On 8/13/13 6:42 AM, Leigh Stoller wrote:

Change the tunnel type to egre-tunnel instead of gre-tunnel. I think that will do the trick.

The reason for this is that openvz and xen cannot talk to each other via GRE tunnels since they are using different encap styles, and I wanted to make sure that we could avoid that case.

On 8/13/13 7:18 AM, Luisa Nevers wrote:

I replaced the link type and re-ran the test, but one of the nodes failed to come up. The slice is "lngre" and the node that failed is pc5.instageni.gpolab.bbn.com Port 35130

On 8/13/13 8:08 AM, Leigh Stoller wrote:

Go ahead and recreate the BBN side of this ...

On 8/13/13 8:47 AM, Luisa Nevers wrote:

Re-created the BBN part of the sliver, but could not ping the Utah host.

Strangely, when I recreated the sliver, the host was ready in less than 2 minutes, which is unusual. Most of my slivers with Xen hosts take about 20 minutes for the host to be ready? Should I just re-create the sliver?

On 8/13/13 8:54 AM, Leigh Stoller wrote:

Two minutes is the right number ... There was another issue related to the older racks missing some DB state.

On 8/13/13 9:01 AM, Luisa Nevers wrote:

Two minutes sound much better, I will capture sliver creation timings for tests I run today.

Should the changed in the DB state affect the current slice? Failing pings went from "From 38.104.187.117 icmp_seq=80 Destination Host Unreachable" to "From vl209.mag02.bos01.atlas.cogentco.com (38.104.187.117) icmp_seq=9 Destination Host Unreachable" but still fail, should I re-create the slivers?

On 8/13/13 9:06 AM, Leigh Stoller wrote:

Should the changed in the DB state affect the current slice?

No, thats a real bug which I will have fixed later today. Well,

But leave the slice active please ...

[lnevers@bbn.com]

This problem is resolved, closing ticket.