Getting watchdog error if in crease the flows to 4000

[jayakumarramisetty-zz]

Code : trex_lib.generate_stateless_udp_streams_per_subnet( c1, port_0, '0c:42:a1:6a:a7:24', '0c:42:a1:6a:a8:3c', '20.1.2.1', '20.1.2.100', '20.1.2.101', '20.1.2.200', 16001, 16158, 1001, 1158, pkt_size=8192, vlan_id=2, vlan_id_incr=1, stream_type='cont', stream_count=4000, total_pps=10000, pkts_per_burst=16 )

trex_lib.generate_stateless_udp_streams_per_subnet( c2, port_1, '0c:42:a1:6a:a8:3c', '0c:42:a1:6a:a7:24', '20.1.2.101', '20.1.2.200', '20.1.2.1', '20.1.2.100', 18001, 18158, 1001, 1158, pkt_size=8192, vlan_id=2, vlan_id_incr=1, stream_type='cont', stream_count=4000, total_pps=10000, pkts_per_burst=16 )

Input config file:

• port_limit : 2 version : 2 #List of interfaces. Change to suit your setup. Use ./dpdk_setup_ports.py -s to see available options interfaces : ['5e:00.2', 'dummy'] #interfaces : ["enp94s0f1_0", "dummy"] port_info : # Port IPs. Change to suit your needs. In case of loopback, you can leave as is. - dest_mac : [0xe4,0x11,0x22,0x33,0x44,0x60]

src_mac : [0x1a,0xfb,0xab,0xae,0x7f,0xad]

        src_mac   : [0xe4,0x11,0x22,0x33,0x44,0x50]
        #src_mac   : [0xe4,0x11,0x22,0x33,0x44,0x50]

port_info : # Port IPs. Change to suit your needs. In case of loopback, you can leave as is.

- dest_mac : [0x98,0x03,0x9b,0x85,0x56,0x3e]

src_mac : [0x98,0x03,0x9b,0x85,0x54,0x06]

platform: master_thread_id: 0 latency_thread_id: 18 dp_flows : 40000000 active_flows : 40000000 dual_if: - socket: 1 threads: [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17] memory: mbuf_9k : 50000

Error on the Trex server while triggered the stateless traffic flow:

*** WATCHDOG violation detected on task ‘Trex DP core 1’ which have failed to response to the signal *** ./t-rex-64: line 100: 23564 Aborted ./_$(basename $0) $INPUT_ARGS $EXTRA_INPUT_ARGS Killing Scapy server... Scapy server is killed

the server is running fine when I am doing for 1000 flows

the core which is getting this watchdog error is random every time among the 18 cores, the server has totally 36 cores and memory available is around 100 gb.

Please help with the inputs or pointer to solve this issue with Trex for the testing purpose

[hhaim]

@jayakumarramisetty this is a design issue. You don't need to add more flows to the trex_cfg. For a single profile we need a single long operation to compile (because we assume that there could be a dependency) and this triggers the watchdog. A workaround for this, you can split you 4000 stream profile to smaller sub-profiles with 1000 streams and add number of profiles. We will look if we can optimize the way we load a single profile (when there is no dependency betwean the streams)

[jayakumarramisetty-zz]

hi haim,

thanks for the reply on this and inputs.

for 4000 flows I understand you suggested for sub profiles, do you have any example for the same if you can provide any example profile that would be helpful.

Also in our test plan we have to do for 1million flows with all different sizes of packets can we do that same with trex (is that doable with tax) if yes please provide if you have any workaround or example for the same that would be so great

[hhaim]

@jayakumarramisetty for generating 1M flows it is better to use Field Engine that can scale better (with one stream you can do it)