machinfo - wron CPU information

[Duncan Edmonstone]

> I've done this before (show_patches -a; swlist)
> The same software, the same patches


are you completely sure?

run sysdiff between the vPars - this will show up *any* software differences

sysdiff sys1 sys2

HTH

Duncan

[Duncan Edmonstone · Accepted solution]

OK, let me take a *guess* at whats causing this...

remember that in vPars we can split CPU cores between partitions... so its possible to have the 1 core with 1/2 the cache in one vPar, and the other core with the other 1/2 of cache in the other vPar.

Now look at the list of CPUs in different vPars:

vPar1:
Boot processor [Path]: 0.121
Monitor assigned [Path]: 0.122
0.123
0.124
1.120
1.121
1.122
1.123
2.120
2.121
2.122
2.123
3.120
3.121
3.122
3.123

vPar2:
Boot processor [Path]: 0.120
Monitor assigned [Path]: 0.125
0.126
0.127
1.124
1.125
1.132
1.133
2.124
2.125
2.132
2.133
3.124
3.125
3.132
3.133

vPar3:
Boot processor [Path]: 5.120
Monitor assigned [Path]: 5.121
5.124
5.125
7.120
7.121

Note that *only* vPar 3 has the other core of the boot processor in the same vPar. (5.120 and 5.121) - So when the macinfo runs, it only gets a report on 1/2 of the resources of the boot processor (it wouldn't surprise me if machinfo made the reasonable assumption that all other processors in the vPar are like the boot processor) It would be interesting to see what you got if you moved these processors around...

Like I said - this is a guess...

HTH

Duncan

[Yarek]

Sorry for delay. I don't have a constant access to this machine.

#sysdiff vPar1 vPar3

KRB5-Client
vPar1 vPar3 description
B.11.31 B.11.31 Kerberos V5 Client Version 1.3.5.03
*PHSS_36361 PHSS_37666 KRB5-Client Version 1.3.5.03 Cumulative patch

krb5client
vPar1 vPar3 description
N/A E.1.6.2 Kerberos V5 Client Version 1.6.2


Machinfo -v and cstm outputs in the attached file.

regards,
Jarek

[Torsten.]

I don't know how machinfo exactly works, but since the cores of the boot processor are splitted and assigned to different vPars, I think this tool only detects the cache that belongs to one of the cores (12MB). The same reason would affect the CPU count. So I agree to Duncan's theory.
In fact all of your CPUs are the same.

Some notes about "Splitting sibling CPUs":

http://docs.hp.com/en/T1335-90083/ch06s24.html

[Yarek]

Thanks guys for the interesting analysis!


regards,
Jarek