dperf 24.11目前支持client使用bond测试吗？

[Laplacence]

RT

[pengjianzhang]

支持

[Laplacence]

感谢回复，我是直接使用test/bond内的测试模版吗？因为里面只有server的模版。

[pengjianzhang]

能不能写个bond的测试文章，可以挂到dperf.org上

[Laplacence]

可以的，同时我也会放到我的blog上：https://laplacence.github.io/

[pengjianzhang]

我就放个你的文章链接

[Laplacence]

现在遇到一些问题

# ../dpdk/usertools/dpdk-devbind.py -s

Network devices using kernel driver
===================================
0000:01:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=3 if=enp1s0f0np0 drv=mlx5_core unused=vfio-pci 
0000:01:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=3 if=enp1s0f1np1 drv=mlx5_core unused=vfio-pci 
0000:25:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=2 if=enp37s0f0np0 drv=mlx5_core unused=vfio-pci 
0000:25:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=2 if=enp37s0f1np1 drv=mlx5_core unused=vfio-pci 
0000:61:00.0 'Ethernet Controller 10G X550T 1563' numa_node=0 if=enp97s0f0 drv=ixgbe unused=vfio-pci *Active*
0000:61:00.1 'Ethernet Controller 10G X550T 1563' numa_node=0 if=enp97s0f1 drv=ixgbe unused=vfio-pci 
0000:81:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=7 if=enp129s0f0np0 drv=mlx5_core unused=vfio-pci 
0000:81:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=7 if=enp129s0f1np1 drv=mlx5_core unused=vfio-pci 
0000:a1:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=6 if=enp161s0f0np0 drv=mlx5_core unused=vfio-pci 
0000:a1:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=6 if=enp161s0f1np1 drv=mlx5_core unused=vfio-pci 


# ./build/dperf -c test/bond/bond.conf 
line 17: error

mode            client

tx_burst                    128
launch_num                  10
cpu                         0

payload_size                1
duration                    1m

cps                         1m
cc                          10m
keepalive                   60s


#port           pci                                  addr         gateway
#               802.ad policy3+4
port            Northbond(0000:25:00.0,0000:25:00.1)  10.11.100.115  10.11.100.1 82:c2:29:82:70:6f

client          10.11.100.115	        1

#               addr_start          num
server          10.11.2.118        1
#               port_start          num
listen          80                  1

提示line 17错误，port这一块我是需要将所有子成员总线地址写上吗？

[Laplacence]

调换了下参数明白了，Northbond需要换成bond4:2，这里的2是指后面两个slave网卡吗？

[pengjianzhang]

参考一下

https://dperf.org/doc/html/configuration#section9

BOND: The format is bondMode:Policy(PIC0,PCI1,…), Mode is [0-4], Policy is [0-2].

[Laplacence]

policy这一块貌似没找到说明

[Laplacence]

# ./build/dperf -c test/bond/bond.conf
not enough ports available: avail 0 require 1
port init fail
dpdk init fail

在server遇到了这个问题，但我的环境是mlnx的cx5，是dpdk识别失败了吗？

# ./../dpdk/usertools/dpdk-devbind.py -s

Network devices using kernel driver
===================================
0000:01:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=3 if=enp1s0f0np0 drv=mlx5_core unused=vfio-pci
0000:01:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=3 if=enp1s0f1np1 drv=mlx5_core unused=vfio-pci
0000:25:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=2 if=enp37s0f0np0 drv=mlx5_core unused=vfio-pci
0000:25:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=2 if=enp37s0f1np1 drv=mlx5_core unused=vfio-pci
0000:61:00.0 'Ethernet Controller 10G X550T 1563' numa_node=0 if=enp97s0f0 drv=ixgbe unused=vfio-pci *Active*
0000:61:00.1 'Ethernet Controller 10G X550T 1563' numa_node=0 if=enp97s0f1 drv=ixgbe unused=vfio-pci
0000:81:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=7 if=enp129s0f0np0 drv=mlx5_core unused=vfio-pci
0000:81:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=7 if=enp129s0f1np1 drv=mlx5_core unused=vfio-pci
0000:a1:00.0 'MT27800 Family [ConnectX-5] 1017' numa_node=6 if=enp161s0f0np0 drv=mlx5_core unused=vfio-pci
0000:a1:00.1 'MT27800 Family [ConnectX-5] 1017' numa_node=6 if=enp161s0f1np1 drv=mlx5_core unused=vfio-pci

[pengjianzhang]

出现了很多tcp reset，socket error；检查一下网络中有没有IP重复

[Laplacence]

#client
seconds 25                 cpuUsage 0   
pktRx   1,759              pktTx    1,756              bitsRx   851,504            bitsTx  817,216            dropTx  0                 
tcpRx   1,756              tcpTx    1,756              udpRx    1                  udpTx   0                 
arpRx   1                  arpTx    0                  icmpRx   0                  icmpTx  0                 
tosRx   0                  otherRx  1                  badRx    0                 
synRx   448                synTx    860                finRx    448                finTx   448                rstRx   412        rstTx 0         
synRt   0                  finRt    0                  ackRt    0                  pushRt  0                 
tcpDrop 0                  udpDrop  1                  ackDup   0                 
skOpen  860                skClose  860                skCon    0                  skErr   0                  rtt(us) 65.5      
tcpReq  448                tcpRsp   448               
ierrors 0                  oerrors  0                  imissed  0  

#server
seconds 25                 cpuUsage 0
pktRx   1,238              pktTx    1,236              bitsRx   600,832            bitsTx  563,616            dropTx  0
tcpRx   1,236              tcpTx    1,236              udpRx    0                  udpTx   0
arpRx   2                  arpTx    0                  icmpRx   0                  icmpTx  0
tosRx   0                  otherRx  0                  badRx    0
synRx   412                synTx    412                finRx    412                finTx   412                rstRx   0          rstTx 0
synRt   0                  finRt    0                  ackRt    0                  pushRt  0
tcpDrop 0                  udpDrop  0                  ackDup   0
skOpen  412                skClose  412                skCon    0                  skErr   0
tcpReq  412                tcpRsp   412
ierrors 0                  oerrors  0                  imissed  0

这次测试应该正常了，之前失败应该是我的两台设备有做bond，一台是8口bond4一台是4口bond4。
测试的时候没有关闭其他端口，导致流量灌入了错误的端口内。

[pengjianzhang]

期待你的文章

[pengjianzhang]

为什么要用Bond呢，dperf可以配置多个port

[Laplacence]

我们这边线上设备是一个bond4的LB，所以想基于这个形态进行测试。