perfSONAR User Q&A and Other Discussion

[Sal Ghani <>]

Doh! Thanks Mark! The omit switch was leftover from when I was running the iperf3 tests, I don’t think it warrants making changes to the plugin unless the community wants it. I just redeployed the psconfig without the omit flag, waiting on results to start coming back in.

 

Sal Ghani Systems/Network Engineer Lonestar Education and Research Network   Mailing Address: PO Box 16920, Lubbock, TX 79490-6920 E-Mail: Telephone: 469-318-0820 | LEARN NOC: 866-647-8728 Website: www.tx-learn.net

 

 

From: Mark Feit <>
Date: Tuesday, July 6, 2021 at 9:01 AM
To: Sal Ghani <>, "" <>
Subject: Re: Psconfig to run iperf2 tests

 

Sal Ghani writes:

 

I’m running a mesh with a number of perfSONAR nodes on which I’d like to run throughput tests using iperf2. My truncated config file looks like this:

 

Test:

"throughput_test":

        {

            "type":"throughput",

            "spec":

            {

                "source":"{% address[0] %}",

                "dest":"{% address[1] %}",

                "duration":"PT30S",

                "omit":"PT5S"

                }

}

 

I’ve checked some of the nodes and I am noticing entries similar to this in their psconfig-pscheduler-agent.log files:

WARN pid=20171 prog=perfSONAR_PS::PSConfig::PScheduler::Agent::_run_end line=231 guid=6FFF77C4-DDB6-11EB-A3F0-965CEB9615E8 msg=Problem adding test throughput(<IP Address of node 1>-><IP Address of node 2>), continuing with rest of config: UNPROCESSABLE ENTITY: Unable to complete request: No tool in common among the participants:  <IP address of node 1> offered iperf3, ethr; < IP address of node 2> offered iperf3, ethr.

 

Iperf2 doesn’t have an omit switch like iperf3, so its plugin will decline to participate when pScheduler shops a test around that has it.  Since the task specifies iperf2 as the sole tool it will accept, there’s nothing available to run it.

 

We could change the plugin so it does the equivalent of omit with what it gets back from iperf2.  My only reservation is the we’d have to force a reporting interval to a value the user might not want and then recalculate some of the aggregate values and report them as if they were iperf2’s numbers.

 

Manually running pscheduler task --tool iperf2 throughput -d <IP address> against the same nodes produces test results as expected. Not quite sure what I’m doing wrong in the psconfig or why the nodes report that there is no tool in common when the manual tests run without issue using the same tool.

 

Add --omit PT5S to the command line and you’ll see the same error:

 

$ pscheduler task --tool iperf2 throughput -d other-host --omit PT5S

Submitting task...

Failed to post task: Unable to complete request: No tool in common among the participants:  localhost offered iperf3, ethr;  other-host offered iperf3, ethr.

 

Hope that helps.

 

--Mark