yahoo-eng-team team mailing list archive

[ibm-cloud-qa <ibm-cloud-qa@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

We have a Juno based openstack cloud build using Packstack. Controller
node, network node and computes nodes each are running on a bare metal
servers with LBaaS agent running on the network node.

We are noticing a low throughput of 1G - 1.2G with HAproxy on a 10G
link.


We are measuring the throughput using iperf tool for the HAProxy. We
have client on one compute node and the servers on different compute
nodes which are linked via 10G . We changed the MTU size from 1400 to
9000 but received the same throughput.

client ---- 10G ----> HAproxy ----- 10G ---> servers


root@instance-00000098:/# iperf -c 20.1.1.200  -i 1 -t 5 -M 1400               
WARNING: attempt to set TCP maximum segment size to 1400, but got 536
------------------------------------------------------------
Client connecting to 20.1.1.200, TCP port 5001
TCP window size: 22.0 KByte (default)
------------------------------------------------------------
[  3] local 20.1.1.108 port 54438 connected with 20.1.1.200 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0- 1.0 sec   120 MBytes  1.01 Gbits/sec
[  3]  1.0- 2.0 sec   140 MBytes  1.17 Gbits/sec
[  3]  2.0- 3.0 sec   138 MBytes  1.16 Gbits/sec
[  3]  3.0- 4.0 sec   135 MBytes  1.13 Gbits/sec

Please let me know if we can tweet any parameter to increase the
throughput.

** Affects: neutron
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1498288

Title:
  Low throughput with haproxy

Status in neutron:
  New

Bug description:
  We have a Juno based openstack cloud build using Packstack. Controller
  node, network node and computes nodes each are running on a bare metal
  servers with LBaaS agent running on the network node.

  We are noticing a low throughput of 1G - 1.2G with HAproxy on a 10G
  link.


  We are measuring the throughput using iperf tool for the HAProxy. We
  have client on one compute node and the servers on different compute
  nodes which are linked via 10G . We changed the MTU size from 1400 to
  9000 but received the same throughput.

  client ---- 10G ----> HAproxy ----- 10G ---> servers

  
  root@instance-00000098:/# iperf -c 20.1.1.200  -i 1 -t 5 -M 1400               
  WARNING: attempt to set TCP maximum segment size to 1400, but got 536
  ------------------------------------------------------------
  Client connecting to 20.1.1.200, TCP port 5001
  TCP window size: 22.0 KByte (default)
  ------------------------------------------------------------
  [  3] local 20.1.1.108 port 54438 connected with 20.1.1.200 port 5001
  [ ID] Interval       Transfer     Bandwidth
  [  3]  0.0- 1.0 sec   120 MBytes  1.01 Gbits/sec
  [  3]  1.0- 2.0 sec   140 MBytes  1.17 Gbits/sec
  [  3]  2.0- 3.0 sec   138 MBytes  1.16 Gbits/sec
  [  3]  3.0- 4.0 sec   135 MBytes  1.13 Gbits/sec

  Please let me know if we can tweet any parameter to increase the
  throughput.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1498288/+subscriptions