[Linux-PowerEdge] Dell R210-II - Unable to disable hyperthreading?

matt matt at theratshack.net
Thu Sep 24 13:45:10 CDT 2015

Hello Linux Poweredge group:

For the past month, I have been tasked with a special problem across a 
worldwide fleet of Dell R210-II servers.  These servers are all running 
Centos 6.7 and until recently have been working well.  We started 
noticing an issue with processor load but were able to successfully 
trace it back to the fact that some servers have hyperthreading enabled 
and some did not.  Further reviews indicated that the HT enabled servers 
were encountering the load issues and that the HT disabled servers were 
not.  The decision was made by management to disable HT on all servers 
fleet-wide in order to mitigate the load issue and I have been tasked 
with doing just that.

I've set out changing HT enablement with some success (10 out of 14 
servers were able to have HT disabled using omconfig) however these last 
four have me scratching my head.  I can't figure out why these four 
servers won't allow me to change HT to disabled.  While most of the test 
group was on BIOS 1.4.0, one of the failed servers was also on v1.4.0 
and the other three were on v1.1.4.

All four servers show the same symptoms:

1) The command "./omconfig chassis biossetup attribute=LogicalProc 
setting=Disabled passwd=********************" results in the error 
"Error! The operation was unsuccessful because the attribute cannot be 
configured using Server Administrator." (This appears to be "working" 
syntax for v1.4.0 BIOS servers)

2) The command "./omconfig chassis biossetup attribute=cpuht 
setting=disabled" results in the error "Error! The specified attribute 
name is not valid. See command help for valid attribute name(s)." (This 
appears to be "working" syntax for v1.1.1 BIOS, although these are rare 
in the fleet.  Only one server with this BIOS version was seen in the 
test group and the command worked successfully.)

3) The command "./omconfig chassis biossetup -?" shows no properties 
whereas all the "working" servers showed all of the available properties 
able to be configured with omconfig.

4) The output of the omconfig commands does not change if OSMA v8.2.0 or 
v8.1.0 is used (although I've upgraded OSMA on the four non-working 
servers to 8.2.0 just to be sure).

5) Comparing dmidecode output on the four affected servers with the 
"good" servers reveals no significant changes relating to BIOS or 
hardware except where expected (RAM manufacturers, processors, hard 
drives, etc..)

According to dsu --inventory, the only difference is the BIOS versions.  
I haven't had a BIOS v1.1.4 success yet, and BIOS v1.4.0 is successful 
except for one failure so I can't figure out the commonality between the 
failures.  I can't just simply say "BIOS v1.1.4 is bad and v1.4.0 is good".

1. OpenManage Server Administrator  ( Version : 8.2.0 )
2. BIOS  ( Version : 1.1.4 )   ** OR ** BIOS  ( Version : 1.4.0 )
3.  iDRAC  ( Version : )
4. NetXtreme BCM5720 Gigabit Ethernet PCIe (eth0)  ( Version : 7.10.17 )
5. NetXtreme BCM5720 Gigabit Ethernet PCIe (eth1)  ( Version : 7.10.17 )

If anyone can shed some light on why I can't get these four servers to 
switch HT, I would be very grateful. If there is additional information 
I can provide to assist with data collection, please let me know.

Thank you in advance!


More information about the Linux-PowerEdge mailing list