Red Hat Bugzilla – Bug 489578
Individual cpu core will fail cpuscaling depending upon what type of drive is used and whether INFO runs before or after test
Last modified: 2014-03-25 20:56:27 EDT
Created attachment 334692 [details]
passing results where cpuscaling is run by itself. No info run.
Description of problem: Core 0 of the cpuscaling test will fail when the drive is a SSD and info is run before the test.
Version-Release number of selected component (if applicable): RH 5.3 x86 or x64 / HTS 5.3-15
How reproducible: Every time
Steps to Reproduce:
1.Install RH 5.3 on a blade that contains an SSD.
2.Load HTS 5.3-15 and create the testplan. Add the cpuscaling test for each core manually. Disable all other tests except for info and cpuscaling.
3.Start test with "hts certify". Info will run and cpuscaling will launch immediately after info completes. Core 0 will fail, the other cores will pass.
Actual results: When info is run first and then cpuscaling, the 0 core will fail every time. If I run cpuscaling without info, it passes every time. If I execute "hts certify --test cpuscaling --device cpu0", info will run after the cpuscaling test and the test will pass.
Expected results:Test passes no matter what drive is used or how the test is launched.
Additional info:We have run the same test substituting a sata driver for the SSD. It will pass no matter what the run order is.
Created attachment 334695 [details]
Failing result where info is run first.
Created attachment 334696 [details]
passing sata results where info was run first.
Looking at the sosreport it appears that this is a Nehalem-based system you're testing on. Does this system support TurboBoost and is it enabled in the BIOS?
Issue is seen with both turboboost on and off.
We were unable to reproduce this issue with RH5 U4 and hts -17
Created attachment 397684 [details]
patch to check loadave is < 1 and perform a sync before starting scaling test
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.