Bug 750686 - cpuscaling on Westmere-EX run too much time
Summary: cpuscaling on Westmere-EX run too much time
Keywords:
Status: CLOSED NOTABUG
Alias: None
Product: Red Hat Hardware Certification Program
Classification: Retired
Component: Test Suite (tests)
Version: 1.4
Hardware: i386
OS: Linux
medium
medium
Target Milestone: ---
: ---
Assignee: Greg Nichols
QA Contact: Guangze Bai
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-11-02 03:22 UTC by Guangze Bai
Modified: 2015-02-08 21:37 UTC (History)
4 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2012-01-11 22:25:45 UTC


Attachments (Terms of Use)

Description Guangze Bai 2011-11-02 03:22:49 UTC
Description of problem:

cpuscaling took too much time and finally the beaker job watchdogged on a Westmere-EX machine with RHEL5.7 i386 arch.

https://beaker.engineering.redhat.com/recipes/310933
http://beaker-archive.app.eng.bos.redhat.com/beaker-logs/2011/10/1491/149141/310933/3410206/TESTOUT.log

https://beaker.engineering.redhat.com/recipes/308845
http://beaker-archive.app.eng.bos.redhat.com/beaker-logs/2011/10/1481/148133/308845/3386579/TESTOUT.log

Version-Release number of selected component (if applicable):
v7-1.4-36

How reproducible:
Not sure. Above 2 beaker jobs produced the issue on the same machine intel-sunriseridge-02.lab.bos.redhat.com. Seemingly often.

Steps to Reproduce:
1.
2.
3.
  
Actual results:


Expected results:
cpuscaling can succeed.

Additional info:
I'll try other Westmere-EX machines.

Comment 1 Caspar Zhang 2011-11-02 06:42:21 UTC
WSM-EX systems has too many CPUs, thus it's not a bug, I just increased the testing time.

Comment 2 Rob Landry 2011-11-02 15:10:47 UTC
The package data looks like rubbish on that box.  It shows every core as being in it's own package which is incorrect.  I've seen -xen do this, but the baremetal kernel should not.

<snip>
package 210 has cpus: 27
package 194 has cpus: 19
package 178 has cpus: 29
package 192 has cpus: 3
package 114 has cpus: 30
package 130 has cpus: 17
package 64 has cpus: 2
package 66 has cpus: 18
<snip>

E7-4870 is a 10 core with hyperthreading or 20 cores in one package.  This also caused it to fail the tests as the package cores need to be changed in sync to have the desired effect.

Comment 3 Caspar Zhang 2011-11-03 11:28:53 UTC
Re-open it... do some more investigation and see which component is suitable for this bug...

seen on intel-sunriseridge-02.lab.bos.redhat.com, RHEL5 i386
https://beaker.engineering.redhat.com/recipes/315218
https://beaker.engineering.redhat.com/recipes/310933

RHEL5 x86_64 seems have correct packages:
CPU  0 Model:         Intel(R) Xeon(R) CPU E7- 4870  @ 2.40GHz
System has 80 cpus
package 1 has cpus: 78, 74, 70, 66, 62, 58, 54, 50, 46, 42, 38, 34, 30, 26, 22, 18, 14, 10, 6, 2
package 0 has cpus: 76, 72, 68, 64, 60, 56, 52, 48, 44, 40, 36, 32, 28, 24, 20, 16, 12, 8, 4, 0
package 3 has cpus: 79, 75, 71, 67, 63, 59, 55, 51, 47, 43, 39, 35, 31, 27, 23, 19, 15, 11, 7, 3
package 2 has cpus: 77, 73, 69, 65, 61, 57, 53, 49, 45, 41, 37, 33, 29, 25, 21, 17, 13, 9, 5, 1

Comment 4 Caspar Zhang 2011-11-03 11:30:32 UTC
RHEL6 is normal as well:

CPU  0 Model:  Intel(R) Xeon(R) CPU E7- 4870  @ 2.40GHz
System has 32 cpus
package 1 has cpus: 2, 6, 10, 14, 18, 22, 26, 30
package 0 has cpus: 0, 4, 8, 12, 16, 20, 24, 28
package 3 has cpus: 3, 7, 11, 15, 19, 23, 27, 31
package 2 has cpus: 1, 5, 9, 13, 17, 21, 25, 29

Comment 5 Guangze Bai 2011-12-06 09:07:05 UTC
Recently cpuscaling was local watchdogged on intel-s3e37-01.rhts.eng.rdu.redhat.com with RHEL5.8-Server-20111121.0_nfs-i386 baremetal kernel:
https://beaker.engineering.redhat.com/recipes/349303

Also, package data is abnormal:
<snip>
CPU  0 Model:  Intel(R) Xeon(R) CPU E7- 4870  @ 2.40GHz
System has 32 cpus
package 210 has cpus: 27
package 194 has cpus: 19
package 178 has cpus: 29
package 192 has cpus: 3
package 114 has cpus: 30
package 130 has cpus: 17
</snip>

Comment 6 Rob Landry 2011-12-12 17:16:18 UTC
intel-s3e37-01.rhts.eng.rdu.redhat.com has 40 cores according to beaker; likely this is expected behavior as the kernel doesn't have full control of all of the system cores.

Comment 7 Guangze Bai 2011-12-13 03:04:42 UTC
So those package data of intel-sunriseridge-02.lab.bos.redhat.com in comment 0 are also expected and acceptable for lack of full control of all system cores, right? If so, close it NOTABUG?


Note You need to log in before you can comment on or make changes to this bug.