What to do when (essential) cleanups fail?
This MR pipeline failure is caused by failure to perform system clean up to a reasonable state by this scenario - following scenarios then can't reasonably work.
We should handle this somehow. But how?
- more resilient clean ups?
- fail the pipeline and skip the rest of the tests? (this is IIRC what @vbenes proposed for missing
/tmp/testeth0
)
Also, I encountered something similar in a bit more treacherous fashion: scenario leaving behind an active connection while not failing itself. Should we have a check comparing some basic things (say connection names and statuses, maybe NM configuration diff) before scenario and after cleanup and fail the scenario if it leaves the system dirty?