<div dir="ltr"><div><div>On our clusters with pure computation we use always HT off, HT was slowing down scientific calculations.<br>Just I'm interested if any benefit of enabling HT on the Ovirt host?<br></div>At least on my system host-Centos6.5, guest-Centos7VM-LAMPstack with separate CentOS7-MariaDBVM did not show any performance boost if I switch on/off HT on 2650v2.<br></div>a.<br><br><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Oct 29, 2014 at 5:24 PM, Xavier Naveira <span dir="ltr"><<a href="mailto:xnaveira@gmail.com" target="_blank">xnaveira@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">On 10/29/2014 05:15 PM, Daniel Helgenberger wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
<br>
On 29.10.2014 16:44, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 04:29 PM, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 04:06 PM, Daniel Helgenberger wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
<br>
On 29.10.2014 15:57, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 03:07 PM, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 01:26 PM, Daniel Helgenberger wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 29.10.2014 11:48, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 11:47 AM, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
On 10/29/2014 11:40 AM, Daniel Helgenberger wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
<br>
On 29.10.2014 10:21, Xavier Naveira wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi,<br>
<br>
We are migrating our ifrastructure from kvm+libvirt hypervisors to<br>
ovirt.<br>
<br>
Everything is working fine but we're noticing that all the<br>
qemu-kvm<br>
processes in the hypervisors take a lot of CPU.<br>
</blockquote>
Without further details of the workload this is hard tell. One<br>
Reason I<br>
can think of might be KSM [1]. Is it enabled on your cluster(s)?<br>
What is<br>
your mem over-commitment setting?<br>
<br>
Note, IIRC the KSM policy is currently hard coded; it will start at<br>
80%<br>
host mem usage.<br>
<br>
[1] <a href="http://www.ovirt.org/Sla/host-mom-policy" target="_blank">http://www.ovirt.org/Sla/host-<u></u>mom-policy</a><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
The typical example is an idle machine, running top from the<br>
machine<br>
itself it reports cpu use percentages below 10% and loads (with 2<br>
processors) of 0.0x. The process running that machine in the<br>
hypervisor<br>
rports cpu uses in the order of the 80-100%.<br>
<br>
Should the values look like this? Why are the idle machines eating<br>
up so<br>
much CPU time?<br>
<br>
Thank you.<br>
Xavier<br>
<br>
______________________________<u></u>_________________<br>
Users mailing list<br>
<a href="mailto:Users@ovirt.org" target="_blank">Users@ovirt.org</a><br>
<a href="http://lists.ovirt.org/mailman/listinfo/users" target="_blank">http://lists.ovirt.org/<u></u>mailman/listinfo/users</a><br>
<br>
</blockquote>
<br>
</blockquote>
<br>
</blockquote>
Hi, thank you for the answer.<br>
<br>
I've been trying to work out some pattern and realized that the VMs<br>
using that much cpu all are Redhat 5.x, the Readhat 6.x doesn't<br>
exhibit<br>
this kind of high cpu use. (we run only redhat/centos 5.x/6.x on the<br>
cluster)<br>
</blockquote>
What OS are the hosts running? In case of EL6, make sure you have<br>
tuned-0.2.19-13.el6.noarch installed [1].<br>
</blockquote>
<br>
That's exactly the version we've in the hypervisors.<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
To further investigate please post Engine, VDSM, libvirt and kernel<br>
versions from the hosts.<br>
</blockquote>
<br>
vdsm-xmlrpc-4.14.11.2-0.el6.<u></u>noarch<br>
vdsm-cli-4.14.11.2-0.el6.<u></u>noarch<br>
vdsm-python-4.14.11.2-0.el6.<u></u>x86_64<br>
vdsm-4.14.11.2-0.el6.x86_64<br>
vdsm-python-zombiereaper-4.14.<u></u>11.2-0.el6.noarch<br>
<br>
libvirt-client-0.10.2-29.el6_<u></u>5.12.x86_64<br>
libvirt-0.10.2-29.el6_5.12.<u></u>x86_64<br>
libvirt-lock-sanlock-0.10.2-<u></u>29.el6_5.12.x86_64<br>
libvirt-python-0.10.2-29.el6_<u></u>5.12.x86_64<br>
<br>
2.6.32-431.23.3.el6.x86_64 #1 SMP Wed Jul 16 06:12:23 EDT 2014 x86_64<br>
x86_64 x86_64 GNU/Linux<br>
<br>
<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
[1] <a href="https://access.redhat.com/solutions/358033" target="_blank">https://access.redhat.com/<u></u>solutions/358033</a><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
I'll take a look to the KSM config.<br>
<br>
Cheers,<br>
<br>
Xavier<br>
<br>
</blockquote>
<br>
</blockquote>
<br>
</blockquote>
<br>
Actually, this seems to be it. But I'm already at a newer kernel:<br>
<a href="https://bugzilla.redhat.com/show_bug.cgi?id=705082" target="_blank">https://bugzilla.redhat.com/<u></u>show_bug.cgi?id=705082</a><br>
</blockquote>
Well, I do not have such hardware so I never run into the issue. You<br>
could disable HT as I suspect your physical cores are less then 64?<br>
<br>
Your workload might differ but my VMs usually do not benefit from<br>
'threaded' cores and I want HT disabled anyway. Also, you can check<br>
cluster settings and disable 'count threads as cores' if enabled. But I<br>
think this might not make any difference.<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
</blockquote>
<br>
</blockquote>
Yeah, these are machines with 4 sockets, 6 core per socket and HT<br>
enabled, so total 48 "CPU".<br>
</blockquote></blockquote>
Good to know; yet the largest host I have has 32 (2 sockets, 8 cores, HT<br>
enabled) CPUs and is not showing this issue (at least I just looked and<br>
everything seems fine).<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
So, are you implying that the problem is the number of "CPUs"? We were<br>
hoping to add some more hypervisors to the cluster next week that have<br>
even more cores...<br>
<br>
I can probably try to disable HT when we add the next hypervisor next<br>
week but it feels that it'd be just a workaround?<br>
</blockquote></blockquote>
Maybe, but not a bad one as you should not have any disadvantages.<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
I opened a bug at redhat just in<br>
case:<a href="https://bugzilla.redhat.com/show_bug.cgi?id=1158547" target="_blank">https://bugzilla.redhat.<u></u>com/show_bug.cgi?id=1158547</a><br>
</blockquote>
I have to ask as I cannot see the BZ because I have no subscription any<br>
more. Against witch component did you open it?<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
</blockquote>
<br>
</blockquote></div></div>
I did as in the original bug: kernel, and then I took KVM as subsystem.<div class="HOEnZb"><div class="h5"><br>
______________________________<u></u>_________________<br>
Users mailing list<br>
<a href="mailto:Users@ovirt.org" target="_blank">Users@ovirt.org</a><br>
<a href="http://lists.ovirt.org/mailman/listinfo/users" target="_blank">http://lists.ovirt.org/<u></u>mailman/listinfo/users</a><br>
</div></div></blockquote></div><br></div>