PDA

View Full Version : dpctl: opening management socket: No such file or directory



duongnt
08-04-2010, 04:13 PM
Hi,
I'm a newbie:). I installed openflow 1.0.0 and netfpga 2.1.1 in centOS5.4 (real machine). everything is good.
in another machine, I installed NOX (0.8beta) in Vmware.
I want to connect netfpga to NOX controller.
I have the two problems :
1, [root@heodaohoa gui]# dpctl adddp nl:0
dpctl: opening management socket: No such file or directory

OpenFlow1.0.0 don't have kernel module. Is this my problem??

2, Can you guide me detailly how to connect NOX to netfpga?( I set bridged between VMware (NOX) and real XP).

I used cable cat5e and eth port in the NOX machine is gigabit port (onboard).
my problem is that two machines can't ping each other:(
please help me!
thanks
ps: sorry . my E is so bad :''>

Yabetatsuya
08-04-2010, 04:58 PM
Hi,
1.
Your suspicion is correct.
See 'run your switch' section of the following OpenFlow wiki to find out how you can set it up:
http://www.openflowswitch.org/wk/index.php/CentOS_NetFPGA_Install
2.
I assume you have a separate machine for NOX.
If you connect them in your local network, one suggestion is that connect onboard port of NetFPGA machine and onboard port of your NOX machine, with or without typical switch (switching-hub), and set both sides(real OS) IP addresses in a same network domain. (I hope this answered to your question.)

Thank you,
Tatsuya

duongnt
08-05-2010, 12:25 PM
Thanks for your quick reply,
Due to your guide, netfpga_openflow connected to NOX controller. :)

Before now, my mistake is that I set for nf2c0 (ifconfig nf2c0 192.168.x.x) before run the commands (in the link which you send to me). so I get error:
[root@localhost ~]# Aug 05 18:39:07|00001|datapath|ERR|nf2c0 device has assigned IP address 192.168.49.2

I got success with the following steps:
1, connect onboard port of NetFPGA machine and onboard port of NOX machine
2. NetFpga machine: ifconfig eth0 192.168.49.70
3. NOX machine: ifconfig eth0 192.168.49.73, XP (static IP 192.168.49.72)
4. NOX machine: cd ~/nox/build/src/
- ./nox_core -i ptcp:
5. NetFPga machine:
- ofdatapath punix:/var/run/test -i nf2c0 &
- ofprotocol unix:/var/run/test tcp:192.168.49.73 --out-of-band &

I got the results below:


root@localhost ~]# ofdatapath punix:/var/run/test -i nf2c0 &
[1] 2885
[root@localhost ~]# RTNETLINK answers: No such file or directory

[root@localhost ~]# ofdatapath punix:/var/run/test -i nf2c0 &
[2] 2911
[root@localhost ~]# ofprotocol unix:/var/run/test tcp:192.168.49.73
Aug 05 17:40:43|00001|secchan|INFO|OpenFlow reference implementation version 1.0.0
Aug 05 17:40:43|00002|secchan|INFO|OpenFlow protocol version 0x01
Aug 05 17:40:43|00003|secchan|WARN|new management connection will receive asynchronous messages
Aug 05 17:40:43|00004|rconn|INFO|unix:/var/run/test: connecting...
Aug 05 17:40:43|00005|rconn|INFO|tcp:192.168.49.73: connecting...
Aug 05 17:40:43|00006|rconn|INFO|unix:/var/run/test: connected
Aug 05 17:40:43|00007|port_watcher|INFO|Datapath id is 002320094054
Aug 05 17:40:43|00008|port_watcher|INFO|Identified data path local port as "tap1".
Aug 05 17:40:43|00009|rconn|INFO|tcp:192.168.49.73: connected
Aug 05 17:40:43|00001|datapath|WARN|unknown vendor: 0x2320

(I have to press ctrl-C if I want to do other commands)



and in NetFPGA machine, wireshark can capture OFP messages.
There are some types of OFP message: Hello, echo, request,...but there is a type is Error??
These messages apear repetitively.

I don't know whether my setup is really correct or not??? I lacked of something???

I have one more question: I want to send something (text, image, audio, video,...) between two machine on the ofdatapath. What I have to do???
Thanks,

Yabetatsuya
08-06-2010, 07:30 AM
Hi,

From the log it seems good except you need to use all the four NetFPGA ports as OpenFlow ports.
Please check that you are using NOX 0.6 and you have checked out openflow-1.0 branch.

As for data connection, OpenFlow parses up to L4 port numbers. So you should be able to decide the actions (which packets are sent to where) from those headers. See OpenFlow spec for details. It's on OpenFlow website.
I also recommend you go through the nox documentation on 'noxrepo' website. Although the description is targeted for OpenFlow 0.8.9, it is the best starting point.

Thanks,
Tatsuya Yabe

duongnt
08-07-2010, 08:38 AM
Hi,
As you see, in my last post, there is a line:
Aug 05 17:40:43|00001|datapath|WARN|unknown vendor: 0x2320
I use wireshark to capture packets, and there are the following packets :
...
- Hello (SM)
- Hello (SM)
- Features Request (CSM)
- Set config (CSM)
- Feature reply (CSM)
- Vendor (SM) (24B)
- Error (SM) (36B)
- Flow Mod (CSM)
- Echo Request (SM)
- Echo Request (SM)
- Echo Request (SM)
- Echo Reply (SM)
- Echo Reply (SM)
- Echo Request (SM)
- Echo Reply (SM)
.............
Then, Echo Request and Echo Reply appear repetitively. Finally, Wireshark stop with the message: Network adapter on which the capture was being done is no longer running, the capture has stopped. ( while Netfpga and NOX connected, I didn't send anything between the two PC)

The Error packet I got above like that (showed in Wireshark):

.......... OpenFlow Protocol:
... Error message:
- Type: Request was not understood (1)
- Code: Vendor not supported (in ofp-vender or ofp-stats-request or ofp-stats-reply)
- Data: 0104...000..232000...000

hic, there are something wrong in my setup? Did I miss something or some step???

Yabetatsuya
08-09-2010, 11:05 PM
Hi,
The error message is no problem. Nox sends one vendor extension message in the first stage.
Since reference_switch doesn't know the vendor specific message, it sends back an error.

For your another issue that wireshark stops, attaching your log file would be helpful.

Thank you,
Tatsuya Yabe

duongnt
08-10-2010, 10:21 AM
Thanks,
1, when run : ofdatapath punix:/var/run/test -i nf2c0,nf2c1,nf2c2,nf2c3 &
sometimes I get the message: RTNETLINK answers: No such file or directory

if it's a error??

I only want to sure that my setup don't have any problem before I deploy a larger testbed:)

2. I don't know if there is a GUI or some other tools to test and flow statistics, ... . I used java GUI in <netfpga_directory>/ lib/java/gui/router.sh without configuration. and I didn't get any packet on the GUI :(.
can u tell me how I have to do with that GUI?? or some tools I can use to test, analyse .... my network with openflow-netfpga switches and NOX controller.

3. I want to deploy some algorithms in data center . should I install RipCord (maybe this question is out of this thread, but I expect to your answer:).

Many thanks,:)

Yabetatsuya
08-10-2010, 04:33 PM
Hi,
1. When you mentioned 'sometimes', do you know in what situations you see the message?
2. For switch: there's no GUI for it. For controller: Please see 'OpenFlow GUI' section of OpenFlow wiki below to find out it is useful for you.
http://www.openflowswitch.org/wk/index.php/Main_Page
3. I cannot say from the info above, but if you have questions about its deployment or topology, 'openflow-discuss' mailing list is a good place to ask (you may want to be more specific):
http://www.openflowswitch.org/wp/people/

duongnt
08-10-2010, 11:46 PM
every time I start netFPGA PC and run ofdatapath punix:/var/run/test -i nf2c0,nf2c1,nf2c2,nf2c3 & , I always get the message
RTNETLINK answers: No such file or directory.
but it didn't appear anymore in a working session.
It means that I only get that message in the first time I run the command ofdatapath (in each working session).

however, the switch connected to the controller without any problem.

*
hi, I got many useful information from your guide:).

thanks so much,

Yabetatsuya
08-11-2010, 06:41 PM
Hi,
Sorry I haven't experienced this...
But from your comment, I understand that it happens every first time you reboot your machine.
If so, before you do nf_download, can you run:
/sbin/ifconfig
and find out if you can see nf2c0,1,2,3 are up? if not, run:
for i in `seq 0 3`; do /sbin/ifconfig nf2c$i up; done
and do the rest of your process.
Please let me know if you still see the issue.
Thanks,
Tatsuya