Hide Forgot
Description of problem: Automated testing environments need to have consistent tests running every time. Rteval warnings like failure to run hackbench stress due to low memory, might go undetected but they impact the accuracy of the results.There should be an option to trigger a failure in case of warnings. How reproducible: Always Steps to Reproduce: 1.Run rteval when there is less than 1Gb alloted per vcpu 2.Hackbench test does not run because of low memory 3.Other stresses/loads continue to run. Actual results: Warning is generated. Expected results: The test should have failed because of low memory.
I have no problems adding a --fail-on-warn option in the NFV/7.4 release timeframe.
I see that in the particular case of hackbench, if there is an OSError due to not enough memory, hackbench is sleeping and waiting to see if some memory is freed up, which might be okay in a server environment, but not what we want in a low-memory environment. This might be more the source of your problem, so I'm investigating whether there is a solution other than the proposed one.
Created attachment 1339246 [details] Remove hack that sleeps if out-of-memory launching hackbench
It appears that this is of no help in a server environment either, so the solution is to remove this code and let rteval fail with an OSError due to not enough memory. Fixed in 2.14-4 and up
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:0969