Hello,
Just to sum up and ask for help since I'm doing a thesis and have stuck
here for a long time. You guys don't have to follow the other 2 emails
because I sum all here. Sincerely thanks for your support.
Experimental mininet 2.1.0 network of 6 ovs switches 2.1.2 and some
hosts (with ip in 192.168.1.0/24) and try to deploy Netflow on the
switches. The controller is opendaylight.
/sudo mn --switch ovsk --controller remote --custom /smile/ncstestmap.py
--topo ncstestmap --arp/
/sh ovs-vsctl -- set Bridge s1 netflow=@nf -- --id=@nf create NetFlow
targets=\"192.168.1.2:5566\" active-timeout=10/
Try ping, udp client and server, but netflow is not working. I installed
from apt-get so most likely I'm working in userspace mode. I tried
compiling kernel module, but the switch crashes when I install netflow
command. (ovs-vswitchd crashed with SIGSEGV in xlate_actions_() ). I ran
the testsuites before install and all tests are successful.
My questions:
1 - If the netflow configuration is wrong, please point out for me. Or
if my concept of "what a flow is" is wrong, please. :)
2 - Do netflow require a kernel module or just user space is enough?
3 - If a kernel module is required, so can you suggest me with the good
version combination between kernel version, ovs version and gcc version?
I have both 12.04 and 14.04 ubuntu and I can manage to downgrade my
kernel version.
Best regards,
Duy
On 07/27/2014 04:01 PM, Duy Bach Ha wrote:
Hello,
An update on the matter, I discovered that the cause of my crashing is
not from connecting to the controller, but because of this command:
/sh ovs-vsctl -- set Bridge s1 netflow=@nf -- --id=@nf create NetFlow
targets=\"192.168.1.2:5566\" active-timeout=10/
I'm on ovs 2.1.2, linux kernel version 3.11 and is working with kernel
module I compiled with gcc 4.7. At first I thought it is because of
the controller, but actually it's not. The moment I add NetFlow
configuration, the switch crashed, than come back again, then crash
again, infinitely. However, in the testsuite, tes-netflow just run
successfully.
/2014-07-26T14:23:31.315Z|00017|daemon(monitor)|WARN|9 crashes: pid
6215 died, killed (Segmentation fault), core dumped, waiting until 10
seconds since last restart//
//2014-07-26T14:23:41.315Z|00018|daemon(monitor)|ERR|9 crashes: pid
6215 died, killed (Segmentation fault), core dumped, restarting/
This is syslog:
/Jul 27 15:35:48 LilBao ovs-vsctl: ovs|00001|vsctl|INFO|Called as
ovs-vsctl -- set Bridge s1 netflow=@nf -- --id=@nf create NetFlow
"targets=\"192.168.1.2:5566\"" active-timeout=10//
//....//
//Jul 27 15:35:49 LilBao kernel: [ 241.781970] ovs-vswitchd[1738]:
segfault at 10 ip 0000000000431469 sp 00007fff8fb8c500 error 4 in
ovs-vswitchd[400000+141000]//
//....//
//Jul 27 15:35:50 LilBao ovs-vswitchd: ovs|00002|daemon(monitor)|ERR|1
crashes: pid 1738 died, killed (Segmentation fault), core dumped,
restarting//
//.... and then after that the log of the restart of all switches/
So again, I hope you guys can give me suggestion on my 3 questions
that I previously asked:
1 - If the netflow configuration is wrong, please point out for me. Or
if my concept is wrong, please. :)
2 - Do netflow require a kernel module or just user space is enough?
3 - If a kernel module is required, so can you suggest me with the
good version combination between kernel, ovs and gcc?
Thank you so much,
Duy
On 07/26/2014 05:16 PM, Duy Bach Ha wrote:
Hi all,
I'm running an experimental mininet network of 6 ovs switches and
some hosts (with ip in 192.168.1.0/24) and try to deploy Netflow on
the switches. The controller I'm using is opendaylight. So for example:
/sudo mn --switch ovsk --controller remote --custom
/smile/ncstestmap.py --topo ncstestmap --arp/
/sh ovs-vsctl -- set Bridge s1 netflow=@nf -- --id=@nf create NetFlow
targets=\"192.168.1.2:5566\" active-timeout=10/
Everything went fine, I can check with ovs-vsctl list NetFlow or list
Bridge and the configuration is there.
My purpose is just to use wireshark to see on (for example) all s1
interfaces the Netflow (cflow in wireshark) packets coming to the
destination. I first try it with ping and then with udp client and
server. All shows up on wireshark. (According to wiki, netflow v5
also consider ping as a flow). But no netflow report is coming out.
In /var/log/openvswitch/ovs-vswitchd.log, no collector failure or
something like that is showing.
I have tried downgraded to kernel version to 3.11.10 (with headers
installed) and compiled a kernel module for 2.1.2 with gcc 4.7 (which
is the one used to compile 3.11.10 for ubuntu). But the switches just
crashed at the moment it connect to the controller. Sometimes at the
first run after compilation work, it works and the next restart it
does not.
/2014-07-26T14:23:31.315Z|00017|daemon(monitor)|WARN|9 crashes: pid
6215 died, killed (Segmentation fault), core dumped, waiting until 10
seconds since last restart//
//2014-07-26T14:23:41.315Z|00018|daemon(monitor)|ERR|9 crashes: pid
6215 died, killed (Segmentation fault), core dumped, restarting/
I search harder and haven't get any idea what is wrong with my
configuration. So here is my questions:
1 - If the netflow configuration is wrong, please point out for me.
Or if my concept is wrong, please. :)
2 - Do netflow require a kernel module or just user space is enough?
3 - If a kernel module is required, so can you suggest me with the
good version combination between kernel, ovs and gcc?
Best regards,
Duy
--
-------------------
Bach Ha Duy
INFOTECH Master Program
University of Stuttgart, Germany
--
-------------------
Bach Ha Duy
INFOTECH Master Program
University of Stuttgart, Germany
Filderbahnplatz 31, 70567 Stuttgart/Mohringen
Flat 01 Room No. 03
0152 59757763
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss
