SD-WAN

I have multiple nodes  on a physical switch using VAN SDN Controller version 2.2.5.

I have the nodes pinging across the SDN instance which respond to each other OK.

The problem I am encountering is that when I execute the REST API

https://10.3.3.103:8443/sdn/v2.0/net/arps

 The resulting json is empty:

{
  "arps": []
}

Is this normal behavior, or should I expect to see arp table entries returned?

Thanks,

Dale Smith

HP SA

Hello Dale Smith,

Could you please check the arp.age setting on the controller UI under com.hp.sdn.ctl.nodemgr.impl.NodeManager ?
As per the release notes for controller v2.2      http://h20565.www2.hp.com/portal/site/hpsc/template.BINARYPORTLET/public/kb/docDisplay/resource.process/?spf_p.tpst=kbDocDisplay_ws_BI&spf_p.rid_kbDocDisplay=docDisplayResURL&javax.portlet.begCacheTok=com.vignette.cachetoken&spf_p.rst_kbDocDisplay=wsrp-resourceState%3DdocId%253Demr_na-c04219762-1%257CdocLocale%253Den_US&javax.portlet.endCacheTok=com.vignette.cachetoken     [ page 6]–
Controller loses knowledge of hosts on the network, but those hosts are actively communicating
through or with the network infrastructure (CR148453)—

You can change the arp.age as suggested in the release note.You will see all learned end host ARPs in the response of the REST API however once the time crosses the timeout value, you see the empty set.

Please try with some new value of arp.age and let us know if you still face the issue.

Thanks,
HP SDN Team

I tried the suggested changes to the ARP timer and saw the same results.  I also cleared ARP caches on both the end station and the gateway router and still did not get an entry in the json structure returned by the API.

Could you suggest a test case so that I can validate my lab configuration?

Thanks,

Dale Smith

Hello Dale Smith,

Did you try with curl command as below-

$ curl -sk -H "X-Auth-Token:$token" https://10.3.3.103:8443/sdn/v2.0/net/arps

Here token already exported .

Are you able to run other REST API (e.g. datapaths,nodes,… ) and see the response?

Please provide details about your switch/config/network topology.

Thanks,

HP SDN Team

Hello Dale Smith,

Doing a follow up to check if you need any further assistance with respect to the query posted.
Please do let us know  if your problem is solved.

If you have more questions on the same topic please do reply on the same thread or open a new post if new topic.


Thanks
HP SDN Team

Yes sorry for the delayed response.

When I create a mininet topology, the API returns a json list of ARP bindings.  This was shortly after a controller restart.

Afterwards, the same REST query returns an empty set.

I will try to see if I can consistently replicate this.

Can you advise which system logs will point to the NodeManager/ARP process status?

Thanks,

Dale Smith

You can increase the arp age. The default value is 5 (in minutes).
Steps:

Click in the General >> Configuration (left side)
Click in com.hp.sdn.ctl.nodemgr.impl.NodeManager >> Modify
Increase the arp.age value

I have changed the arp timers with no noticeable effects.  I have also cleard arp cahces on all devices attached to the OpenFlow instance with no change, with the exception of mininet as noted in my previous post.

I also note that I never get arp response (json structure) from a "real" switch (HP 3800 with Nashville release).

However, functionally the OpenFlow instance performs as expected with end points able to arp neighbors within the broadcast domain.  I have a pyhsical gateway router attached to the OpenFlow instance and it shows correct arp bindings for the neighbors that are active.

If someone can suggest a log that I can view to determine if there is an issue with my controller runtime environment or other test, that would be appreciated.

Hello Dale Smith

Could you please check the following to see if the end hosts are discovered:

1. /net/nodes:  should list the end hosts
2. /net/links:  should list the links between the end hosts (paths configured by controller)
3. Change ‘Debug’ log level for node manager
   1. Connect to http://localhost:8443/admin/content/logging from the controller itself.  Don’t forget that the username/password is admin/sdn unless the you have changed that in /opt/sdn/virgo/configuration/org.eclipse.virgo.kernel.users.properties
   2. Change log level to ‘Debug’ in ‘hp.sdn.net.node’.
   3. Now run pings and see if you get something like the following in the log file located in the /var/log/sdn/virgo/logs/access directory:

[2014-07-07 16:01:05.892] DEBUG of-io-37-thread-5            hp.sdn.net.node                                                   Node 00:15:60:ae:87:19-10.250.200.130 discovered off 00:c8:00:1c:2e:e7:e4:00 on port 0x29.

Please let us know if this helps.

Thanks,

HP SDN Team