Thanks, Robie! Our users reported that issues usually appear within one or two hours, but that's likely tied to machine-specific workloads.
I've been able to force this issue to occur in a VM when running the test script from the description together with stress-ng memory/vm stressors. The "failed to read" logs then trigger consistently within a few minutes.
Considering that we do need memory pressure for the ENOBUFS issues to happen, I think this is a valid test scenario. I'll add this to the test plan section of the bug, and see if I can do a long duration test with the package from proposed.
Thanks, Robie! Our users reported that issues usually appear within one or two hours, but that's likely tied to machine-specific workloads.
I've been able to force this issue to occur in a VM when running the test script from the description together with stress-ng memory/vm stressors. The "failed to read" logs then trigger consistently within a few minutes.
Considering that we do need memory pressure for the ENOBUFS issues to happen, I think this is a valid test scenario. I'll add this to the test plan section of the bug, and see if I can do a long duration test with the package from proposed.