Bug 1062661

I have been sent a similar issue as per below. But this time its on a Citrix XenServer. 

As up to RHEL 6.3 (kernel-2.6.32-220.el6.x86_64) the subscription-manager command would return the following:
 
+-------------------------------------------+
    Installed Product Status
+-------------------------------------------+
 
ProductName:            Red Hat Enterprise Linux Server
Version:                6.3
Arch:                   x86_64
Status:                 Subscribed
Starts:                 08/01/2012
Expires:                08/01/2015
 
As of RHEL 6.4 (kernel-2.6.32-358.18.1.el6.x86_64) the subscription-manager command would return the following:
 
+-------------------------------------------+
    Installed Product Status
+-------------------------------------------+
Product Name:   Red Hat Enterprise Linux Server
Product ID:     69
Version:        6.4
Arch:           x86_64
Status:         Partially Subscribed
Status Details: Only covers 2 of 8 sockets.
Starts:         08/01/2012
Ends:           08/01/2015
 
Our current pool’s description is as follows:
 
Subscription Name: Red Hat Enterprise Linux Server, Standard (1-2 sockets) (Unlimited guests)

(In reply to Ken Hitchcock from comment #6)
> I have been sent a similar issue as per below. But this time its on a Citrix
> XenServer. 
> 

> Status:         Partially Subscribed
> Status Details: Only covers 2 of 8 sockets.


I suspect this (incorrect socket count on Xen) is more likely related to https://bugzilla.redhat.com/show_bug.cgi?id=844532


This bug (1062661) seems to be ppc46 LPAR specific.

I think this is the same issue as https://bugzilla.redhat.com/show_bug.cgi?id=1070908

ppc64 lpar systems add a /sys/devices/system/cpu/cpu* entry for each cpu thread available to the physical system (regardless if they are associated with that lpar). Code was previously assuming a cpu* entry meant a cpu available to the system (similar to other arches). But that is not true for ppc64 lpar systems.

The fix for 1070908 is to ignore cpu* entries that do not provide any cpu topology information. With that change, the count of cpu threads drops to 8, and
the topo info is calculated as:


'cpu.core(s)_per_socket' : '1'
'cpu.cpu(s)' : '8'
'cpu.cpu_socket(s)' : '2'
'cpu.thread(s)_per_core' : '4'
'cpu.topology_source' : 'kernel /sys cpu sibling lists'


Now, this is where things get complicated. For what the kernels considers a 'socket', that is a 2 socket machine. IBM, however, considers that a single socket (by a different definition of socket) machine.

Can RH provide the patch for this and also a test package?

IBM is signed up to test and provide feedback, setting OtherQA.

Hello IBM,
Please confirm if you would be able to run:
# yum update subscription-manager
on RHEL 6.6 install.

Then test if this issue has been corrected.

Thank You
Joe Kachuck

Hello IBM,
This should be fixed in subscription-manager-1.12.14-7.el6 and above.

Thank You
Joe Kachuck

As stated in the comments above, this issue should already be addressed by the fix for bug 1070908 and is available in subscription-manager-1.11.2-1 and newer.  RHEL 6.6 was released with subscription-manager-1.12.14-7.el6 which includes the fix.  Please re-open if the problem continues after running...

yum update subscription-manager
subscription-manager facts --update