perfSONAR User Q&A and Other Discussion

[Brian Candler <>]

On 16/11/2017 20:34, Andrew Lake wrote:
> If there was an outage meaning it couldn’t even talk to OWAMPD on the 
> other side at port 861 then you won't get any test results. If OWAMP 
> can’t even connect to port 861 it can't do things like choose the UDP 
> ports or exchange results, so it would be incorrect to label that as 
> 100% packet loss since it didn't send any packets. The only time you 
> get 100% packet loss is if the OWAMP client can connect to port 861 on 
> the remote end to establish a session and get the UDP ports for the 
> test, and then it loses all the UDP packets (usually a firewall causes 
> that). It’s possible we could display these gaps better, but the 
> reporting is working as expected in the face of a hard failure like 
> you describe.

Thanks for the explanation.  However one thing is unclear: if the sides 
were initially able to establish an owamp control session on port 861 
and started exchanging packets, and logging the test results every 60 
seconds, but then the packets were suddenly to stop coming in - why is 
this not noticed?  Does the local side depend on receiving a report from 
the far side at the end of each 60 second window?

It would be helpful if failed owamp session could produce a red dot for 
"failed", in the same way that failed throughput tests do.

Not being able to see outages reduces the usefulness of the tool.  I 
guess I could write some code which scrapes esmond and looks for gaps 
longer than 60 seconds between the packet loss entries.

Regards,

Brian.