VMware Cloud Community
MarkB1979
Contributor
Contributor
Jump to solution

ESXi Detecting more CPUs than are physically present.

I have just installed a new Cisco UCS C240 server with 2 physical CPUs (verfied via CIMC & Physical inspection).  I installed ESXi 5.5 only for it to inform me that the servere has 4 CPUs in it.  This is physically impossible however now Vcentre will not allow me to apply the avalible 2CPU license that I have spare.  I have attempted to contact Vmware support however that was a abject waste of my time as they palmed me off to the "licensing team" who spent an hour looking at it and telling me what I already knew that it needed to go to the technical team. 

Has anyone else seen this issue where ESXi is reporting more physical CPUs that are actually physically present?

Thanks,

Mark

1 Solution

Accepted Solutions
Alistar
Expert
Expert
Jump to solution

Hi there,

I think I have seen a thread with the exact same issue some time ago - basically there is a feature that spans the CPU across 2 NUMA nodes - this should be a performance-enhancing feature but is not yet supported by 5.x version of ESXi (not sure about 6.x though). You need to see the processor settings in BIOS/UEFI and check the CPU settings - you should find a new feature listed there -  disable it Smiley Happy

Stop by my blog if you'd like 🙂 I dabble in vSphere troubleshooting, PowerCLI scripting and NetApp storage - and I share my journeys at http://vmxp.wordpress.com/

View solution in original post

0 Kudos
4 Replies
Alistar
Expert
Expert
Jump to solution

Hi there,

I think I have seen a thread with the exact same issue some time ago - basically there is a feature that spans the CPU across 2 NUMA nodes - this should be a performance-enhancing feature but is not yet supported by 5.x version of ESXi (not sure about 6.x though). You need to see the processor settings in BIOS/UEFI and check the CPU settings - you should find a new feature listed there -  disable it Smiley Happy

Stop by my blog if you'd like 🙂 I dabble in vSphere troubleshooting, PowerCLI scripting and NetApp storage - and I share my journeys at http://vmxp.wordpress.com/
0 Kudos
MKguy
Virtuoso
Virtuoso
Jump to solution

If you're using Haswell-based Xeons (E5 v3 CPUs) then this is normal, see this thread with the same issue:

https://communities.vmware.com/thread/503089

-- http://alpacapowered.wordpress.com
vNEX
Expert
Expert
Jump to solution

Hi,

just in addition to already mentioned ... Cisco have introduced on some UCS systems a feature called Cluster on Die (COD).

Cisco Unified Computing System BIOS Settings - Cisco

Cluster on Die (CoD) snoop is available on Intel Xeon processor E5-2600 v3 CPUs that have 10 or more cores. Note that some software packages do
not support this setting: for example, at the time of writing, VMware vSphere does not support this setting. CoD snoop is the best setting to use when NUMA is
enabled and the system is running a well-behaved NUMA application. A well-behaved NUMA application is one that generally accesses only memory
attached to the local CPU. CoD snoop provides the best overall latency and bandwidth performance for memory access to the local CPU. For access to remote
CPUs, however, this setting results in higher latency and lower bandwidth. This snoop mode is not advised when NUMA is disabled.

This feature is nicely clarified in this article:

http://www.enterprisetech.com/2014/09/08/intel-ups-performance-ante-haswell-xeon-chips/

New feature called Cluster On Die, or COD, that breaks the Haswell Xeon E5 chips with 10 or more cores into different NUMA regions to make sure the data doesn’t

drift too far from the cores in these beasts. It essentially breaks the chip at the interconnect switch and treats it like two chips as far as the software is concerned, and

for many workloads, this actually boosts performance significantly.

Important part is that this feature is not supported by VMware ... below is support statement:

VMware KB:     Known limitations of the Intel® Xeon® processor E5 v3 series with ESXi 5.1.x and ESXi...

ESXi 5.1 with patch ESXi510-201407001 and ESXi 5.5 Update 2 do not support the Cluster-On-Die feature on Intel Xeon processor E5 v3 series

vSphere does not support the Cluster-On-Die feature on Intel Xeon processor E5 v3 series.
Workaround: ESXi 5.1 with patch ESXi510-201407001 and ESXi 5.5
Update 2 systems must have the Cluster-On-Die feature on Intel Xeon processor E5
v3 series disabled because it is currently not supported by vSphere.

 

From USC Release 2.2(3c) you should be able to suppress COD feature by changing CPU Snoop settings in UCSManager via Service Profile and BIOS policy.

Release Notes for Cisco UCS Software, Release 2.2 - Cisco

Change it to one of these settings:

- Early Snoop (ES) is available on all Intel Xeon processor E5-2600 v3 CPUs. It provides good local memory latency and bandwidth, but with a severe
penalty in bandwidth for remote CPU access. ES should be used for latency-sensitive applications that do not require high remote bandwidth: for
example, certain online transaction processing (OLTP) database workloads.

- Home Snoop (HS) is also available on all Intel Xeon processor E5-2600 v3 CPUs and is excellent for NUMA applications that need to access a
remote CPU on a regular basis. Of the snoop modes, HS provides the best remote CPU bandwidth and latency, but with the penalty of slightly higher local
latency, and is the best choice for memory and bandwidth-intensive applications: for example, certain decision support system (DSS) database workloads.


_________________________________________________________________________________________ If you found this or any other answer helpful, please consider to award points. (use Correct or Helpful buttons) Regards, P.
MarkB1979
Contributor
Contributor
Jump to solution

Thanks folks.  Very much appreciated!  Smiley Happy

0 Kudos