vmtests: `pkg.sensors.tracing:TestKprobeRateLimit` and `NoRateLimit` are failing silently because of `panic` in tests
mtardy opened this issue · comments
I realized that the test was failing while writing the v1.0 backport #2584. Only to realize later that it was also failing on main and v1.1, while the vmtests recap being green (for example see, a v1.1 run or a main run). So it was failing everywhere because nc.openbsd
was missing.
So it seemed that when tests panic, they are failing according to vmtests but not counted toward the total. The only remaining question is: what is the difference between v1.0 and v1.1/main?
So I see a few things:
- Fix the tests to use
Fatal
orFailNow
#2685 - Add
nc.openbsd
to the base image used by vmtests
cilium/little-vm-helper-images#582
Also the test is really slow on success (40s+) because it tries to check for missing events (so it's retrying multiple time, it should be optimized not to make the test suite too slow.