HPE VAN Controller and PICA

[themisAnagno]

Hello everyone,

I am using an HPE VAN SDN Controller Version: 2.8.8.0366 and a PICA8 P3297 switch on my network. I want to create ovs bridges on PICA8 that are controlled by the HPE VAN controller. The OVS version on the PICA8 switch is 2.9.1.2/e9c9eaf.  Unfortunately, the controller is able to negotiate the OpenFlow version with the switch, but it cannot get any further information. Any help would be much appreceated.

The configuration on the PICA8 is:

vs-vsctl add-br test_br
ovs-vsctl set bridge test_br datapath_type=pica8
ovs-vsctl set-controller test_br tcp:10.30.0.90:6653
ovs-vsctl set bridge test_br protocols=OpenFlow10,OpenFlow13
# Create the external ports
for i in {5..10};
do
ovs-vsctl add-port test_br ge-1/1/$i vlan_mode=access -- set interface ge-1/1/$i type=pica8;
done0

Bridge test_br
Controller "tcp:10.30.0.90:6653"
is_connected: true
Port "ge-1/1/6"
Interface "ge-1/1/6"
type: "pica8"
Port test_br
Interface sixers
type: internal
Port "ge-1/1/9"
Interface "ge-1/1/9"
type: "pica8"
Port "ge-1/1/8"
Interface "ge-1/1/8"
type: "pica8"
Port "ge-1/1/10"
Interface "ge-1/1/10"
type: "pica8"
Port "ge-1/1/7"
Interface "ge-1/1/7"
type: "pica8"
Port "ge-1/1/5"
Interface "ge-1/1/5"
type: "pica8"

The OF messages between the controller and the switch are:

* 301 2018-10-18T09:37:30.066Z Disc 75:e5:48:6e:73:02:00:20 

* 302 2018-10-18T09:37:30.815Z Conn 75:e5:48:6e:73:02:00:20 

* 303 2018-10-18T09:37:30.815Z Rx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,HELLO,8,12050]}

* 304 2018-10-18T09:37:30.815Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,HELLO,16,12050],elems=VERSION_BITMAP}

* 305 2018-10-18T09:37:30.815Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,FEATURES_REQUEST,8,982291]}

* 306 2018-10-18T09:37:30.835Z Rx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,FEATURES_REPLY,32,982291],dpid=75:e5:48:6e:73:02:00:20,#buf=256,#tab=254,aux=0,cap=[flowStats, tableStats, portStats, groupStats]}

* 307 2018-10-18T09:37:30.835Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,SET_CONFIG,12,982292],flags=null,msLen=65535}

* 308 2018-10-18T09:37:30.835Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REQUEST,16,982293],DESC,flgs=null,body=(no body)}

* 309 2018-10-18T09:37:30.835Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REQUEST,16,982294],PORT_DESC,flgs=null,body=(no body)}

* 310 2018-10-18T09:37:30.835Z Tx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REQUEST,16,982295],TABLE_FEATURES,flgs=null,body={TableFeats: count=0}}

* 311 2018-10-18T09:37:30.857Z Rx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REPLY,1072,982293],DESC,flgs=[],body={mfr:Pica8, Inc,hw=P3297,sw=PicOS 2.9.1.2,...}}

* 312 2018-10-18T09:37:30.857Z Rx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REPLY,208,982294],PORT_DESC,flgs=[],body={Array::Ports: count=3}}

* 313 2018-10-18T09:37:30.992Z Rx 75:e5:48:6e:73:02:00:20 {ofm:[V_1_3,MULTIPART_REPLY,64000,982295],TABLE_FEATURES,flgs=[replyMore],body={TableFeats: count=38}} 

The log messages on the HPE VAN Controller are:

* today 12:35:57.043

INFO

hp.of.ctl

of-idle-timer

Datapath removed: 75:e5:48:6e:73:02:00:20, neg=V_1_3, ip=10.143.143.100

* today 12:36:28.050

WARN

hp.of.ctl

of-idle-timer

Closing unresponsive connection from 10.143.143.100

[ShaunWackerly]

When I've seen issues like this previously, it's been because the controller pushed down the "delete all flows" flowmod to initialize the OpenFlow tables, and the control-plane connection was crossing the OpenFlow tables. This meant the switch dropped packets on the control-plane, which severed the connection. Would packets on the control-plane cross any data-plane tables in this case?

We may need to capture control-plane packets using tcpdump on the controller. That would let you know about any TCP retries or other messages which lead to the controller thinking it cannot reach the switch.