Some remaining questions on running performance tests about opennetvm HOT 12 CLOSED

Hi @Wasdns
Can you please provide the sudo ./tools/dpdk_nic_bind.py --status output. Maybe the NIC you're trying to use is not supported? Make sure the NICs are not blacklisted in the run-pktgen.sh

For part 2 if I understand you correctly the openNetVM machine will just forward incoming packets to the other machine, that can be done easily using our bridge NF. You could also chain more NFs if you want, but the bridge is the simplest for your setup.

from opennetvm.

Hi, @koolzz Thank you for your answer.

"Can you please provide the sudo ./tools/dpdk_nic_bind.py --status output. Maybe the NIC you're trying to use is not supported? Make sure the NICs are not blacklisted in the run-pktgen.sh"

The output of sudo ./tools/dpdk_nic_bind.py --status:

Network devices using DPDK-compatible driver
============================================
0000:01:00.0 'NetXtreme II BCM5716 Gigabit Ethernet' drv=igb_uio unused=
0000:01:00.1 'NetXtreme II BCM5716 Gigabit Ethernet' drv=igb_uio unused=

Network devices using kernel driver
===================================
<none>

Other network devices
=====================
<none>

Crypto devices using DPDK-compatible driver
===========================================
<none>

Crypto devices using kernel driver
==================================
<none>

Other crypto devices
====================
<none>

The NICs are not in the blacklist of run_pktgen.sh.

"For part 2 if I understand you correctly the openNetVM machine will just forward incoming packets to the other machine, that can be done easily using our bridge NF. You could also chain more NFs if you want, but the bridge is the simplest for your setup."

I followed your instructions and start a service chain which is composed of a bridge and a monitor. And I created a topologic like:

h1 <---> Server which runs openNetVM SC <---> h2

h1 runs scapy client(i.e.sending packets), and in contrast, h2 runs scapy server(i.e.receiving packets). And we applied the experiment here: scapy-test: using scapy to execute openNetVM experiments.

Based on our observation, we conclude these results:

• (1)If we bind the two NIC ports to DPDK, we could not see any packets through wireshark, the monitor console and the console of PC2 server.
• (2)If we not bind the two NIC ports to DPDK, we could see lots of packets through wireshark, however, nothing output on both the monitor console and the console of PC2 server.

I don't know if the DPDK NIC ports would ignore the packets or not. Could you give me some advices on it? Thanks a lot!

from opennetvm.

Hi, @Wasdns
You won't see packets through wireshark when using DPDK NICs. You should see numbers in onvm_mgr and other NFs you have running.

How exactly are you running onvm_mgr?
Are your machines directly connected?

from opennetvm.

Hello, @koolzz
Thank you for your explanations. Exactly, I strictly followed the instructions of openNetVM documents here. The command that I used to run openNetVM manager:

$ ./onvm/go.sh 0,1,2,3,4 -s stdout

I don't know if we should start onvm_mgr with respect to manager instructions or not. Besides, all the machines are directly connected.

Please point out our mistakes if we are wrong. Thank you very much!

from opennetvm.

@Wasdns
I see, the ./go.sh script you're using expects port number argument.
The way you're running it is with ports disabled.

Try $ ./onvm/go.sh 0,1,2,3,4 3 -s stdout for 2 ports

from opennetvm.

@koolzz

Thanks a lot. However, it occurs error when initialing the port 0. The error information like:

EAL: Detected 16 lcore(s)
EAL: No free hugepages reported in hugepages-1048576kB
EAL: Probing VFIO support...
EAL: PCI device 0000:01:00.0 on NUMA socket 0
EAL:   probe driver: 8086:1521 net_e1000_igb
EAL: PCI device 0000:01:00.1 on NUMA socket 0
EAL:   probe driver: 8086:1521 net_e1000_igb
Creating mbuf pool 'MProc_pktmbuf_pool' [27648 mbufs] ...
Creating mbuf pool 'NF_INFO_MEMPOOL' ...
Creating mbuf pool 'NF_MSG_MEMPOOL' ...
Port 0 init ... 
Port 0 socket id 0 ... 
Port 0 Rx rings 1 ... 
EAL: Error - exiting with code: 1
  Cause: Cannot initialise port 0

I tried to debug this bug but found nothing. The output of ./dpdk/tools/dpdk-devbind.py --status:

Network devices using DPDK-compatible driver
============================================
0000:01:00.0 'I350 Gigabit Network Connection' drv=igb_uio unused=
0000:01:00.1 'I350 Gigabit Network Connection' drv=igb_uio unused=

Network devices using kernel driver
===================================
<none>

Other network devices
=====================
<none>

Crypto devices using DPDK-compatible driver
===========================================
<none>

Crypto devices using kernel driver
==================================
<none>

Other crypto devices
====================
<none>

I have used setup_environment.sh to initial the environment and freed enough hugepages. How to resolve this situation? Thanks a lot!

from opennetvm.

Looks like a hugepage issue, can you try the hugepage fixes here

from opennetvm.

@koolzz
I have freed the hugepage and the output of grep -i huge /proc/meminfo:

AnonHugePages:      274432 kB
HugePages_Total:       2048
HugePages_Free:        2048
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB

And I tried the instructions of Troubleshooting:

The manager and NFs are not running, but something crashed without freeing hugepages.
To fix this, please run sudo rm -rf /mnt/huge/* to remove all files that contain hugepage data.

The problem remained. I rebooted the machine and the same thing happened.

from opennetvm.

I think this is a NIC driver issue--if it will run without any ports configured that means the basic system (including huge pages) is working. If the Speed Tester example is working that is also a good sign.

Have you tested whether the basic DPDK examples work (e.g. L2fwd)? We have only tested OpenNetVM with Intel-based 10GB NICs. If the DPDK examples run, then you can try to look at how they initialize ports compared to how we do it.

from opennetvm.

I agree with @twood02.

@Wasdns I see two different NICs listed in your comments. Can you check http://dpdk.org/doc/nics and make sure your NICs are supported by DPDK?

from opennetvm.

@twood02 @nks5295 Thank you for your helps!

"Have you tested whether the basic DPDK examples work (e.g. L2fwd)? We have only tested OpenNetVM with Intel-based 10GB NICs. If the DPDK examples run, then you can try to look at how they initialize ports compared to how we do it."

I create the same topologic as mentioned above. I have implemented the DPDK example "l2fwd" and used scapy to test. The test passed and both the packet receiver and the console which runs l2fwd could successfully reports the infomation, including the number of sent packets and received packets.

I see two different NICs listed in your comments. Can you check http://dpdk.org/doc/nics and make sure your NICs are supported by DPDK?

The first NIC list shows the NIC information of the server which runs Pktgen, and another one shows the NIC information of the server which maintains openNetVM SC. Exactly what we discussed above are two different tests: one for testing openNetVM by using Pktgen, another one for testing openNetVM by using scapy or iperf. I could see the abovementioned NIC types on the list at http://dpdk.org/doc/nics.

from opennetvm.

I don't resolve this problem currently on my server. I leverage another approach to run my experiments and I think I should close this issue. But I will report any process as soon as I try openNetVM. Thank you for your great efforts and kind helps!

from opennetvm.