fix: remove server network interface workaround

[jooola]

Remove a server workaround that was used to fix a private network issue.

The problem that this workaround tried to solve was a timing issue. When the workaround was implemented, the action returned by the attach to network endpoint was marked as success without waiting for the network definition to be applied on the host. Today, the actions returned by the attach to network endpoint do wait for the network definition to be applied on the host.

The workaround can therefor be removed with a certain confidence.

Supersede #1020

Fixes #1019

Related to #552 and #550

[codecov]

Codecov Report

Attention: Patch coverage is 83.33333% with 2 lines in your changes missing coverage. Please review.

Project coverage is 66.28%. Comparing base (93d8b1e) to head (6a1553f).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
internal/testtemplate/template.go	83.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1021      +/-   ##
==========================================
+ Coverage   66.13%   66.28%   +0.15%     
==========================================
  Files          66       66              
  Lines        9874     9868       -6     
==========================================
+ Hits         6530     6541      +11     
+ Misses       2628     2612      -16     
+ Partials      716      715       -1     

Flag	Coverage Δ
e2e	66.28% <83.33%> (+0.15%)	⬆️
unit	30.09% <83.33%> (+0.06%)	⬆️

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.

[apricote]

This issue that caused the workaround only happens sporadically, a single successful test run does not give me confidence that the issue went away.

IIRC we tried to setup some e2e tests that were able to reproduce a few months ago with a bastion server and SSH, not sure if we even merged that.

[samene]

This issue that caused the workaround only happens sporadically, a single successful test run does not give me confidence that the issue went away.

IIRC we tried to setup some e2e tests that were able to reproduce a few months ago with a bastion server and SSH, not sure if we even merged that.

But the workaround causes issues with userdata scripts. If we stop a server which is midway running cloud-init it can cause uncertain behaviour. some scripts do not run again on the next boot.

[apricote]

I understand that you have an issue and we should fix that.

I just wanted to say that the suggested solution of removing the workaround is also not without downsides, as it might introduce another bug.

@jooola We previously tried to remove this workaround in the plugin-framework-server branch and it looks like we at least finished writing the test case that validates the behavior the workaround originally tried to fix: 3c16a22

[jooola]

Marking as ready to review, only to run the e2e.

[jooola]

We found the origin of the workaround. I am confident that we can remove it without breaking anything.

[apricote]

Thank you @jooola for researching the original conditions that lead to the workaround and making sure the underlying problem is now fixed.