[ovs-discuss] Queues not working properly (?)

Rach Man costasisme at gmail.com
Fri Sep 29 14:27:07 UTC 2017

Hello everyone,

I have created a very simple topology of one OVSwitch and 3 hosts in
Mininet (latest version from Github on Ubuntu 16.04 LTS) connected to a
remote OpenDaylight controller. All links are TCLinks with bw=10mbps.

h1 and h2 are sending udp packets (with iperf) to h3, in different ports.
h1->h3 (flow 1) is in queue 1 and h2->h3 (flow 2) is in q0. I understand
that if I set the min-rate correctly flow 1 should not drop under min-rate.
Instead flow 2 should only experience packet losses.

Nonetheless, I see both flows 1 and 2 experience packet losses.

Inside my Python script there's the following:

s1 = net.get('s1')
#I clear all queues and qos from the switch, otherwise every change stays
print s1.cmd('ovs-vsctl -- --all destroy QoS -- --all destroy Queue')
print s1.cmd('sudo ovs-vsctl -- set Port s1-eth3 qos=@newqos -- \
                --id=@newqos create qos type=linux-htb
other-config:max-rate=10000000 queues=0=@q0,1=@q1 -- \
                --id=@q0 create queue  other-config:min-rate=5000000
other-config:max-rate=5000000 -- \
                --id=@q1 create queue  other-config:min-rate=3000000

I test using iperf with:
h1: iperf -u -c -p 40001 -b 2m -t 20, h3: iperf -u -s -p 40001
h2: iperf -u -c -p 40002 -b 9m -t 20, h3: iperf -u -s -p 40002

and get:
flow 1: 4.19 Mbytes 1.66Mbits/sec 0.332 ms 406/3403 (12%)
flow 2: 4.19 Mbytes 8.08Mbits/sec 0.332 ms 736/15314 (4.8%)

ovs-ofctl dump-flows s1 shows:
priority=950, udp, nw_src=,nw_dst=,tp_dst=44441
priority=1000, udp, nw_src=,nw_dst=,tp_dst=44442

Does anybody have an idea why this is happening? I'm quite new to OVS so I
might not be able to see the obvious.

Thank you in advance,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20170929/8c04fee9/attachment.html>

More information about the discuss mailing list