Don't use '-A' in network.py on non-numa systems (BugFix)

[rodwsmith]

Description

This PR fixes bug #570, which can cause the CPU load to spike if the CPU affinity is set incorrectly on systems that don't support NUMA features.

The fix involves identifying systems that lack NUMA support and not passing the -A {nodes} option to iperf3 on such systems.

Resolved issues

Resolves bug #570

Jira card: https://warthogs.atlassian.net/browse/SERVCERT-1183

Documentation

No documentation changes required.

Tests

Tested on lloobee (my own ARM64-based development system), which exhibited the bug. The new version drops the CPU load from 90%+ to about 35-40%, and increases the throughput by about 5% (from 88% to 93%, give or take a percent or two).

I've also tested on two servers in the Server Certification pool, torchtusk and whomp, to be sure the script doesn't create problems on systems that DO support NUMA. No unexpected problems arose. The change should result in exactly the same iperf3 command being used on such systems as the previous version of the script. Such systems were unaffected by the bug to begin with.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (cac5044) 36.58% compared to head (3cda9ca) 35.46%.
Report is 4 commits behind head on main.

❗ Current head 3cda9ca differs from pull request most recent head 7ec6662. Consider uploading reports for the commit 7ec6662 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #711      +/-   ##
==========================================
- Coverage   36.58%   35.46%   -1.12%     
==========================================
  Files         310      301       -9     
  Lines       34639    34254     -385     
  Branches     5965     5934      -31     
==========================================
- Hits        12671    12149     -522     
- Misses      21399    21557     +158     
+ Partials      569      548      -21     

Flag	Coverage Δ
provider-base	5.99% <100.00%> (+0.49%)	⬆️
release-tools	?

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bladernr]

Looks good to me... thanks

[Hook25]

I have to ask a few changes here:

• codecov is complaining because we need a new test here, sumarize_cpu is untested
• it is kind of weird to return -1 instead to throwing an exception. Would you consider changing this to raise an exception both if the file is not found or if it is found and contains a -1? That way anyone calling the function will not have to check if -1 is a valid numa number or what is going on, if something is returned, it is valid, else an exception is thrown
• Minor: Change the title of the PR after rebasing on main, this is a bugfix so add a (bugfix) postfix to the title!

Ty

[kissiel]

I added tests and fixed the PR title.
The -1 as the node index is in line with what sysfs would report on a non-numa systems, so I left it in.

@hook please take another look

[Hook25]

+1, if this causes a problem we can change in the future I guess