Description of problem: As part of the Certification suite, there is a "clock test" that looks for "jitter" on SMP systems. This test gets the value of time across several "cores" and checks to see that the difference between the highest and lowest values is bound within 0.2 secs. We have found that under some circumstances, this test would fail: - if there is CPU stress - if there is a large number of logical CPUs (16, or 32) Our developement feels that in the future, based upon how the jitter calculation is currently done, this test will be more difficult to pass as the number of cores become much larger (64 or 128). Dev comments: "One way to deal with this would be to increase the priority of the process to "RT". We performed this modification and noticed much better results. The test passed in every case." Version-Release number of selected component (if applicable): Steps to Reproduce: 1. Execute core test on a RH5/RH4 system while load is running. 2. Move to a higher cpu core count. 3. Actual results: Expected results: Additional info:
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2010-0365.html