<div dir="ltr">I've tried the kernel-lt with same results.<div>For kernel-ml, the baremetal wasn't able to boot at that time.</div></div><br><div class="gmail_quote"><div dir="ltr">În mar., 23 aug. 2016 la 02:22, Adi Pircalabu <<a href="mailto:adi@ddns.com.au">adi@ddns.com.au</a>> a scris:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Laurentiu,<br>
<br>
Just to chip in, without thoroughly checking the thread (which I'd like<br>
to apologize in advance for): have you tried to test other kernel<br>
instead of the stock one? You can have a look at<br>
<a href="http://elrepo.org/tiki/kernel-ml" rel="noreferrer" target="_blank">http://elrepo.org/tiki/kernel-ml</a> and <a href="http://elrepo.org/tiki/kernel-lt" rel="noreferrer" target="_blank">http://elrepo.org/tiki/kernel-lt</a><br>
and use them for some comparison testing.<br>
Cheers,<br>
<br>
---<br>
Adi Pircalabu<br>
<br>
On 22-08-2016 18:31, Laurentiu Soica wrote:<br>
> No luck with qemu-kvm-ev, the behavior is the same. Running perf<br>
> record -a -g on the baremetal shows that most of the CPU time is in<br>
> _raw_spin_lock<br>
><br>
> Children Self Command Shared Object<br>
> Symbol<br>
> - 93.62%<br>
> 93.62% qemu-kvm [kernel.kallsyms] [k]<br>
> _raw_spin_lock<br>
><br>
><br>
> - _raw_spin_lock<br>
><br>
><br>
> + 45.30% kvm_mmu_sync_roots<br>
><br>
><br>
> + 28.49% kvm_mmu_load<br>
><br>
><br>
> + 25.00% mmu_free_roots<br>
><br>
><br>
> + 1.12% tdp_page_fault<br>
><br>
> În joi, 18 aug. 2016 la 11:59, Laurentiu Soica a<br>
> scris:<br>
><br>
>> I've tried with KSM disabled and nothing changed.<br>
>><br>
>> I've upgraded KVM to qemu-kvm-ev. I'm waiting to see if there are<br>
>> any improvements and report back.<br>
>><br>
>> În mie., 17 aug. 2016 la 15:10, Boris Derzhavets<br>
>> <<a href="mailto:bderzhavets@hotmail.com" target="_blank">bderzhavets@hotmail.com</a>> a scris:<br>
>><br>
>> For myself KSM is unpredictable feature. The problem is Compute,<br>
>> just this node<br>
>><br>
>> does "copy on write" , so only Compute.<br>
>><br>
>> My concern exactly is where would it lead to worse or better Guest<br>
>> behavior ?<br>
>><br>
>> I am not expecting complete fix. I would track via top/htop and<br>
>> dmesg via Cron on 1-2 hr<br>
>><br>
>> period.<br>
>><br>
>> -------------------------<br>
>><br>
>> FROM: <a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a><br>
>> <<a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a>> on behalf of Laurentiu Soica<br>
>> <<a href="mailto:laurentiu@soica.ro" target="_blank">laurentiu@soica.ro</a>><br>
>><br>
>> SENT: Wednesday, August 17, 2016 6:38 AM<br>
>><br>
>> TO: Discussion about the virtualization on CentOS<br>
>> SUBJECT: Re: [CentOS-virt] Nested KVM issue<br>
>><br>
>> Both baremetal and compute ? Are there any other metrics do you<br>
>> consider useful to collect for troubleshooting purposes ?<br>
>><br>
>> În mie., 17 aug. 2016 la 13:04, Boris Derzhavets<br>
>> a scris:<br>
>><br>
>> It sounds weird, but attempt to disable KSM and see would it help or<br>
>> no ?<br>
>><br>
>> -------------------------<br>
>><br>
>> FROM: <a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a><br>
>> <<a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a>> on behalf of Laurentiu Soica<br>
>><br>
>> SENT: Wednesday, August 17, 2016 4:56 AM<br>
>><br>
>> TO: Discussion about the virtualization on CentOS<br>
>> SUBJECT: Re: [CentOS-virt] Nested KVM issue<br>
>><br>
>> Enabled the logging on both compute and baremetal. Nothing strange<br>
>> in logs:<br>
>><br>
>> on baremetal :<br>
>><br>
>> Wed Aug 17 11:51:01 EEST 2016: committed 62310764 free 58501808<br>
>> Wed Aug 17 11:51:01 EEST 2016: 87025667 < 123574516 and free ><br>
>> 24714903, stop ksm<br>
>><br>
>> on compute:<br>
>><br>
>> Wed Aug 17 08:52:52 UTC 2016: committed 24547132 free 76730936<br>
>> Wed Aug 17 08:52:52 UTC 2016: 45139624 < 102962460 and free ><br>
>> 20592492, stop ksm<br>
>><br>
>> and the compute node is again at 100% CPU utilization.<br>
>><br>
>> În mar., 16 aug. 2016 la 15:26, Boris Derzhavets<br>
>> a scris:<br>
>><br>
>> I would enable ksmtuned logging ,if it has been done verify logs<br>
>><br>
>> -------------------------<br>
>><br>
>> FROM: <a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a><br>
>> <<a href="mailto:centos-virt-bounces@centos.org" target="_blank">centos-virt-bounces@centos.org</a>> on behalf of Laurentiu Soica<br>
>><br>
>> SENT: Tuesday, August 16, 2016 7:16 AM<br>
>><br>
>> TO: Discussion about the virtualization on CentOS<br>
>> SUBJECT: Re: [CentOS-virt] Nested KVM issue<br>
>><br>
>> Yes. It is on both baremetal and compute node.<br>
>><br>
>> _______________________________________________<br>
>> CentOS-virt mailing list<br>
>> <a href="mailto:CentOS-virt@centos.org" target="_blank">CentOS-virt@centos.org</a><br>
>> <a href="https://lists.centos.org/mailman/listinfo/centos-virt" rel="noreferrer" target="_blank">https://lists.centos.org/mailman/listinfo/centos-virt</a><br>
> _______________________________________________<br>
> CentOS-virt mailing list<br>
> <a href="mailto:CentOS-virt@centos.org" target="_blank">CentOS-virt@centos.org</a><br>
> <a href="https://lists.centos.org/mailman/listinfo/centos-virt" rel="noreferrer" target="_blank">https://lists.centos.org/mailman/listinfo/centos-virt</a><br>
><br>
> --<br>
><br>
> Laurentiu Soica<br>
> _______________________________________________<br>
> CentOS-virt mailing list<br>
> <a href="mailto:CentOS-virt@centos.org" target="_blank">CentOS-virt@centos.org</a><br>
> <a href="https://lists.centos.org/mailman/listinfo/centos-virt" rel="noreferrer" target="_blank">https://lists.centos.org/mailman/listinfo/centos-virt</a><br>
</blockquote></div><div dir="ltr">-- <br></div><div data-smartmail="gmail_signature"><div dir="ltr">Laurentiu Soica</div></div>