perfSONAR User Q&A and Other Discussion

[Marcos Felipe Schwarz <>]

I think that your suggestion is a better way forward Andrew!

Is there any documentation on how to setup a Elastic backend for a central managed perfsonar deployment?

 

For the short term, our solution was to limit our meshes to different bandwith values...

 

Regards,

Marcos Schwarz

 

De: Andrew Lake <>
Enviada em: quinta-feira, 22 de outubro de 2020 16:02
Para: Tim Chown <>
Cc: Marcos Felipe Schwarz <>;
Assunto: Re: [perfsonar-user] Help configuring similar throughput meshes

 

 

 

 

On October 22, 2020 at 5:13:24 AM, Tim Chown () wrote:

Hi Marcos,

This is a very nice idea for comparing congestion control mechanisms.  If you get it working we’d be very interested to see the measurements and their visualisation.

Andy, perhaps this is something to factor in to the visualisation requirements for 4.4?

For sure. The nice thing about the Elastic stuff is that these types of things are "factored-in" by default since we store all this information and have the ability to leverage the query language through platforms like Grafana to build these on the fly if needed. 

 

 

Tim

> On 21 Oct 2020, at 14:26, Andrew Lake <> wrote:
> 
> Hi Marcos,
> 
> The problem is the way that MaDDash queries Esmond so unfortunately splitting the meshes won’t help (which you observed). In Esmond, it does indeed separate your tests but when MaDDash queries Esmond it asks for “all throughput tests matching a given source, destination, and protocol” by default (it also will check ip version, tool and bandwidth if any of those are explicitly set in your pscheduler test definition). It does not currently filter by congestion control algorithm.
> 
> If you are feeling adventurous, attached is a file I think will get you closer to what you want. It won’t fix the graphs MaDDash displays, but will fix the nagios checks that color the MaDDash boxes. Copy the attached file to /usr/lib/perfsonar/psconfig/checks/ps-nagios-throughput.json and restart psconfig-maddash-agent. That is the file that tells MaDDash how to build the Nagios check. The updated version has some blocks that pull the “congestion” parameter out of the pscheduler test spec and pass it to the generalized —filter parameter in the Nagios check that tells it to look at the “pscheduler-throughput-congestion” field in esmond. I have not tested this but in theory should work.
> 
> Thanks,
> Andy
> 
> 
> On October 21, 2020 at 8:45:59 AM, Marcos Felipe Schwarz () wrote:
> 
>> Hi guys. We are creating two throughput meshes to compare BBR and CUBIC, it seems that both tests are running, but maddash was showing the same results on both meshes.
>> 
>> Is this an unsupported scenario? Is there suggested way around?
>> 
>>  
>> We tried using two different dashboard for each test, but they still show the same information.
>> 
>>  
>> Regards,
>> 
>>  
>> Marcos Schwarz
>> 
>>  
>> Gerente de P&D | R&D Manager
>> 
>> Gerência de Execução de P&D em Ciberinfraestrutura | Management of R&D Execution in Cyberinfrastructure
>> 
>> Diretoria de Pesquisa e Desenvolvimento | Board of Research and Development
>> 
>> RNP - Rede Nacional de Ensino e Pesquisa | Brazilian National Research and Educational Network
>> 
>> Promovendo o uso inovador de redes avançadas | Promoting the innovative use of advanced networks
>> 
>> http://www.rnp.br | +55 (19) 3787- 3386 | Skype ID: marcos.f.sch
>> 
>> Campinas - SP - Brasil | E-mail:
>> 
>>  
>> -- 
>> To unsubscribe from this list: https://lists.internet2.edu/sympa/signoff/perfsonar-user 

> --
> To unsubscribe from this list: https://lists.internet2.edu/sympa/signoff/perfsonar-user