perfSONAR User Q&A and Other Discussion

["Chevalier, Scott S" <>]

Greetings Stack,

Probably could've reinforced this a couple more times during the workshop, 
but you came to the correct conclusion already - these little Brix boxes are 
under-resourced to support the full toolkit build. __ For purposes of the 
training last week it was decided to go full toolkit for familiarity of the 
resources there and ease of install.

Spoke with Andy (@ESnet) this morning and he's likely to get back to you this 
week regarding your meshconfig error. He'll probably also have more 
detail-oriented queries for you regarding the OS build and the firewall.

Ps - top-notch issue-tracking and details. Thanks!

Cheers,

Scott Chevalier
Network Systems Analyst
International Networks @ IU <http://in.iu.edu/>
812-856-9964
 

On 6/2/18, 5:36 PM, 
"
 on behalf of 
"
 
<
 on behalf of 
>
 wrote:

    Greetings,
    
    Yesterday (2018.06.01) I attended the [1] Great Plains Network - FIONA 
Node Workshop to learn how to build perfSONAR nodes. As part of the workshop, 
we did our testing on these [2] tiny GIGABYTE EL-20-3700-32GB devices.
    
    [1] https://conferences.k-state.edu/gpn/schedule-2/workshops/ 
<https://conferences.k-state.edu/gpn/schedule-2/workshops/>
    [2] 
https://github.com/GreatPlainsNetwork/20180601-FIONAworkshop/tree/master/hardware/fionette
 
<https://github.com/GreatPlainsNetwork/20180601-FIONAworkshop/tree/master/hardware/fionette>
    
    I hacked away many hours last night rebuilding, experimenting, and 
breaking it repeatedly with perfSONAR. I've been hackin away at it many hours 
today too. I've come to the conclusion that I just simply can't run a full 
toolkit on this poor little thing. It just cranks away on all four cores and 
I end up pushing 60C on the temp (passively cooled) and then I get all kinds 
of fun errors. :-)
    
    So, I did another rebuild. Rocking the latest version of Scientific Linux 
7.5 and then I did an install of the following:
    perfsonar-toolkit-security
    perfsonar-testpoint
    perfsonar-toolkit-systemenv-testpoint
    perfsonar-toolkit-sysctl
    perfsonar-toolkit-servicewatcher
    perfsonar-toolkit-ntp (Still curious as to why ntp over default of chrony 
but whatever :-)
    Now I can run a bunch of the tests and whatnot via command line and 
things are running SO MUCH better. I tossed this device out into my DMZ, 
poked a few holes in my firewall to route the traffic listed in the docs to 
this node, and successfully tested throughput to several hosts I found in the 
lookup service.  Temp is low and I'm not pegging all the cores when I do 
tests! Awesome!
    
    But I want to see the pretty graphs too...
    
    No worries, I thought. I will just set up a Central Measurement Archive 
(got the idea from the pictures here [3]). Built up a vm in my private 
environment with a bunch of memory and loads of cores running Scientific 
Linux 7.5. Ran through the install [4] page again selecting just the 
perfsonar-centralmanagement package to install then followed the link 
directions to go to the configuration page [5]. Err....what? OK. I get the 
general idea of the mesh...but I don't see how that is relevant in the 
slightest. I am looking to pull the results of the tests I created via cli 
from the node for the pretty graph. I've only got one node and I'm not really 
ready to expand my needs to a mesh yet. Ah. Here we go [6] Deploying a 
Central Measurement Archive. I already installed, scroll down, this is a 
simple setup at my house skip down to "Authenticating by IP Address", then 
commands to run. Bingo!
    
    [3] https://docs.perfsonar.net/install_options.html 
<https://docs.perfsonar.net/install_options.html>
    [4] https://docs.perfsonar.net/install_centos.html 
<https://docs.perfsonar.net/install_centos.html>
    [5] https://docs.perfsonar.net/multi_overview.html 
<https://docs.perfsonar.net/multi_overview.html>[6] 
http://docs.perfsonar.net/multi_ma_install.html 
<http://docs.perfsonar.net/multi_ma_install.html>
    
    (root) $ cd /usr/lib/esmond
    (root) $ . bin/activate
    (esmond) (root) $ python esmond/manage.py add_user_ip_address <My user on 
the node> <IP of the node>
    Traceback (most recent call last):  File "esmond/manage.py", line 6, in 
<module>
        import settings # Assumed to be in the same directory.
      File "/usr/lib/esmond/esmond/settings.py", line 16, in <module>
        raise Error("ESMOND_ROOT not defined in environment")
    NameError: name 'Error' is not defined
    
    
    Err. Uh oh. 
    I've been poking away for 45 minutes trying to get passed this point.
    
    1. Any idea what I'm doing wrong?
    2. Am I just doing the wrong thing here for wanting to just view the node 
tests? 
    3. The way my network traffic flows is that the CMA host can initiate 
traffic to the DMZ host, but not the other way. Do I need to poke holes in my 
firewall to allow the DMZ node to communicate back to the VM or can it just 
be a "pull only" scenario?
    Thanks!
    ~Stack~

Attachment: smime.p7s
Description: S/MIME cryptographic signature