KVM Hypervisor - guest steal time stats on the hypervisor

[timcenty]

Hello,

Does anyone know the CentOS / sysstat pkg version where this patch is available at all ?

https://www.spinics.net/lists/kvm/msg116517.html

I don't see it on CentOS 7.9 with sysstat 10.1.5

Thanks

[BShT]

this is a 2015 bug report

do you have any problem?

[TrevorH]

As far as I can see, sysstat on 7.9.2009 reports steal stats.

Code: Select all

[root@centos7 ~]# mpstat -P ALL 1
Linux 3.10.0-1160.71.1.el7.x86_64 (centos7.trevor.local) 	08/09/22 	_x86_64_	(2 CPU)

03:22:26     CPU    %usr   %nice    %sys %iowait    %irq   %soft  %steal  %guest  %gnice   %idle
03:22:27     all    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00  100.00
03:22:27       0    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00  100.00
03:22:27       1    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00    0.00  100.00

[timcenty]

I don't believe the steal reported in 'mpstat' is from guests. One of the VMs on a KVM hypervisor I've has steal around 15% but the 'mpstat' on the hypervisor always show 0.

As per the patch, the '/proc/pid/stat' and pidstat will show the guest steal information. The kernel version shown there is 3.19 though.

I want to get the guest steal cpu information from the hypervisor.

[TrevorH]

Then it's incredibly unlikely that the patch will ever make it back into (RHEL) CentOS 7 is in its last 18 months or so of maintenance before it goes EOL in 2024. Almost all fixes that come out for el7 now are for security bugs. The days where RH would add functionality are long gone (about 3 years ago). RHEL 7 is an operating system from 2014.

I've also just tested the sample pidstat command from your link on el8 and el9 and neither have steal columns in the output. I am not sure this patch ever went anywhere as it doesn't seem to be in any newer distro.

[timcenty]

Thanks for checking. I checked in RHEL 8.2 but don't see it.

So have to ask the patch provider I think. I will do that and see if I can get any update on that.

Thanks