perfSONAR User Q&A and Other Discussion

[Aaron Brown <>]

Hey Ivan,

If you edit /opt/perfsonar_ps/ls_registration_daemon/etc/ls_registration_daemon.conf, there are 3 blocks, “local_latency_service”, “local_bandwidth_service” and “local_web_service”. The first are for ping, traceroute and owamp endpoints. The middle is for bwctl, and the latter is for the Measurement Archive. Just set the interfaces that you’d like, as needed, and then restart the “ls_registration_daemon” service. It may take a bit to age out the old data.

Cheers,

Aaron

On Nov 6, 2014, at 10:38 AM, Garnizov, Ivan (RRZE) <> wrote:

Hi Aaron, group,

 

I also miss a  perspective in the perfSONAR documentation about how to define which services get advertised where.

Is there a CLI procedure for adjusting this (I can’t seem to find the UI config option for it), since it can be easily blocked through “limits” file or the iptables, but my idea is that I would like this to be announced to the public through the LS as well.

 

So to sum up. We would like to have the measurements separated on 2 interfaces of same host, but also would like to advertise it so through LS.

 

Best regards,

Ivan

 

 

From:  [] On Behalf Of Aaron Brown
Sent: Thursday, November 06, 2014 2:04 PM
To: Pete Siemsen
Cc: 
Subject: Re: [perfsonar-user] are BWCTL and OWAMP scheduled separately in 3.4?

 

Hey Pete,

 

On Nov 5, 2014, at 11:45 AM, Pete Siemsen <> wrote:

perfSONAR 3.4 supports hosts with 2 network interfaces in such a way that you can run bandwidth tests on one interface and latency tests on the other, without conflict. Cool.

 

A buddy attended the recent Technology Exchange conference, where he heard that 3.4 will schedule tests in sequence, so that you don't have to worry about conflicts even if your host has just one interface.

 

The online docs for 3.4.1 still recommend using separate interfaces or hosts. So I think 3.4 doesn't solve the conflicting bandwidth/latency issue.

 

Correct?

 

Right now, the web GUI configuration only supports doing the separate interfaces, not co-scheduling. However, you can manually configure the tests to support co-scheduling the two, though, that’s been less well tested, and would be a bit less accurate. If you’d like to try that out, I can put together some documentation.

 

Cheers,

Aaron