TCP throughput drops to zero every 1 second in a Containerlab SR Linux bridge setup

[lou-lan]

                         Nokia SR Linux (leaf1)
                     +------------------------------+
                     |            leaf1             |
                     |                              |
                     |    e1-1               e1-2   |
                     +------+----------------+------+
                            |                |
                            |                |
                            |                |
                10.0.1.1/24 |                | 10.0.1.2/24
                     MTU 9232                MTU 9232
                          |                        |
                     +----+----+              +----+----+
                     |         |              |         |
                     | host1   |              | host2   |
                     |         |              |         |
                     |  eth1   |              |  eth1   |
                     +---------+              +---------+

$ cat simple-leaf-topo.clab.yml
name: simple-leaf

mgmt:
  ipv4-subnet: 172.20.20.0/24

topology:
  kinds:
    nokia_srlinux:
      image: ghcr.io/nokia/srlinux:25.7

  nodes:
    leaf1:
      kind: nokia_srlinux

    host1:
      kind: linux
      image: debian:12

    host2:
      kind: linux
      image: debian:12

  links:
    - endpoints: ["leaf1:e1-1", "host1:eth1"]
    - endpoints: ["leaf1:e1-2", "host2:eth1"]

host 1

root@host1:/# ip a show eth1
131: eth1@if130: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9232 qdisc noqueue state UP group default
    link/ether aa:c1:ab:0d:c5:da brd ff:ff:ff:ff:ff:ff link-netnsid 1
    inet 10.0.1.1/24 scope global eth1
       valid_lft forever preferred_lft forever
    inet6 fe80::a8c1:abff:fe0d:c5da/64 scope link
       valid_lft forever preferred_lft forever

host2

root@host2:/# ip a show eth1
129: eth1@if128: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9232 qdisc noqueue state UP group default
    link/ether aa:c1:ab:76:78:a4 brd ff:ff:ff:ff:ff:ff link-netnsid 1
    inet 10.0.1.2/24 scope global eth1
       valid_lft forever preferred_lft forever
    inet6 fe80::a8c1:abff:fe76:78a4/64 scope link
       valid_lft forever preferred_lft forever

iperf3 test

root@host1:/# iperf3 --time 60 --zerocopy --client 10.0.1.2
Connecting to host 10.0.1.2, port 5201
[  5] local 10.0.1.1 port 47294 connected to 10.0.1.2 port 5201
[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
[  5]   0.00-1.00   sec   332 KBytes  2.72 Mbits/sec    3   8.96 KBytes.   # first sec look has some speed 2.72 Mbits/sec 
[  5]   1.00-2.00   sec  0.00 Bytes  0.00 bits/sec    1   8.96 KBytes      # <--- i don’t know why speed is 0
[  5]   2.00-3.00   sec  0.00 Bytes  0.00 bits/sec    0   8.96 KBytes
[  5]   3.00-4.00   sec  0.00 Bytes  0.00 bits/sec    1   8.96 KBytes
[  5]   4.00-5.00   sec  0.00 Bytes  0.00 bits/sec    0   8.96 KBytes
[  5]   5.00-6.00   sec  0.00 Bytes  0.00 bits/sec    0   8.96 KBytes
[  5]   6.00-7.00   sec  0.00 Bytes  0.00 bits/sec    1   8.96 KBytes

I’m not sure if this is normal. I don’t really understand the forwarding architecture of nokia/srlinux (maybe the poor performance is caused by user-space forwarding?). I’d like to know how to troubleshoot this.

ping test

root@host1:/# ping 10.0.1.2 -c 3
PING 10.0.1.2 (10.0.1.2) 56(84) bytes of data.
64 bytes from 10.0.1.2: icmp_seq=1 ttl=64 time=0.298 ms
64 bytes from 10.0.1.2: icmp_seq=2 ttl=64 time=0.220 ms
64 bytes from 10.0.1.2: icmp_seq=3 ttl=64 time=0.735 ms

--- 10.0.1.2 ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2053ms
rtt min/avg/max/mdev = 0.220/0.417/0.735/0.226 ms

[hellt]

Hi
SR Linux version < 25.10 have aggregated 1000pps limit across all ports, this is why you see iperf stalling.

In versions >= 25.10 the pps limit is increased to 10 000 pps