[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH] nodeinfo: Add workaround if kernel reports bogous numa topology.



On Tue, Oct 30, 2012 at 10:08 PM, Eric Blake <eblake redhat com> wrote:
Both bitmaps show all 24 cores, so hwloc is able to read sysfs and
determine the existence of 2 sockets with 12 nodes each, and where the
12 nodes are numbered 0-5 twice according to which bank of cache they
are tied to.  Which version of libvirt is this tested on where libvirt
was only reporting 12 cores, because I thought we already patched that
with commit 80533ca in 0.10.0.  That is, I think proc1.png should result in:

$ virsh nodeinfo
    CPU model:           x86_64
    CPU(s):              24
    CPU frequency:       2200 MHz
    CPU socket(s):       2
    Core(s) per socket:  12
    Thread(s) per core:  1
    NUMA cell(s):        1
    Memory size:         8047272 KiB


(Just for clarity, I am the original reporter, sorry I couldn't answer earlier, I subscribed to the list too late and didn't want to break the thread)

On the host with 1 NUMA cell, with libvrit 0.10.2 I get:

[root host29 ~]# virsh nodeinfo
CPU model:           x86_64
CPU(s):              24
CPU frequency:       2200 MHz
CPU socket(s):       2
Core(s) per socket:  6
Thread(s) per core:  1
NUMA cell(s):        1
Memory size:         131971548 KiB

Furthermore, to answer the question at the bottom of your email, http://birzan.org/capabilities1.txt has this info.
 
and proc4.png would _ideally_ result in:

$ virsh nodeinfo
    CPU model:           x86_64
    CPU(s):              24
    CPU frequency:       2200 MHz
    CPU socket(s):       2
    Core(s) per socket:  12
    Thread(s) per core:  1
    NUMA cell(s):        4
    Memory size:         8047272 KiB


[root host34 ~]# virsh nodeinfo
CPU model:           x86_64
CPU(s):              24
CPU frequency:       2200 MHz
CPU socket(s):       2
Core(s) per socket:  6
Thread(s) per core:  1
NUMA cell(s):        1
Memory size:         131971020 kB

And http://birzan.org/capabilities4.txt 

This is on 0.9.11.5, on 0.10.2:

[root host34 libvirt]# virsh nodeinfo
CPU model:           x86_64
CPU(s):              24
CPU frequency:       2200 MHz
CPU socket(s):       1
Core(s) per socket:  6
Thread(s) per core:  1
NUMA cell(s):        4
Memory size:         131971020 KiB

and http://birzan.org/capabilities4-newlibvirt.txt 

This has made the problem even worse, as we now can only use 6 cores out of 24 by default (libvirt pins qemus to the CPUs is sees available, so we have to manually taskset them after starting).
 
I think the CPU _is_ reporting the complete NUMA topology through sysfs,
but that we are probably consolidating information from the wrong files
and therefore getting confused.

The sysfs is tarred up in http://birzan.org/sysdevicessystem.tar.gz for host29 (the one with 1 numa cell in capabilities) and http://birzan.org/sysdevicessystem-34.tar.gz
 
Also, what does the 'virsh capabilities' report for the <topology>
section?  Whereas 'virsh nodeinfo' is constrained by back-compat to give
a lame answer for number of NUMA cells, at least 'virsh capabilities'
should be showing a reasonable representation of the machine's topology.


I answered this above. If you need any more info, feel free to ask.
 

--
George-Cristian Bîrzan

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]