Bug 1644693

Summary:	High Host CPU load for Windows 10 Guests (Update 1803) when idle
Product:	[oVirt] ovirt-engine	Reporter:	Michal Skrivanek <michal.skrivanek>
Component:	General	Assignee:	Steven Rosenberg <srosenbe>
Status:	CLOSED CURRENTRELEASE	QA Contact:	Liran Rotenberg <lrotenbe>
Severity:	medium	Docs Contact:	Rolfe Dlugy-Hegwer <rdlugyhe>
Priority:	urgent
Version:	future	CC:	ailan, bugs, chayang, claytonn, darinschmidt, dholler, jinzhao, jspanko, juzhang, knoel, lijin, linux, lspiehler, mailinglists35, mark.dinicola, mavital, michal.skrivanek, michen, mkalinin, mstevens, mtessun, nerijus, pbonzini, pgozart, rbarry, rdlugyhe, ribarry, sbonazzo, siliu, srosenbe, tburke, virt-bugs, virt-maint, vkuznets, vrozenfe, xfu, xiagao, yfu, zhguo
Target Milestone:	ovirt-4.3.0	Keywords:	FutureFeature
Target Release:	---	Flags:	michal.skrivanek: ovirt-4.3? lrotenbe: testing_plan_complete+ rule-engine: planning_ack? rule-engine: devel_ack+ lrotenbe: testing_ack+
Hardware:	x86_64
OS:	Linux
Whiteboard:
Fixed In Version:	vdsm-4.30.5, ovirt-engine-4.3.0_rc	Doc Type:	Enhancement
Doc Text:	Previously, for virtual machines with a Windows 10 guest, the host CPU load was too high. The current release reduces the CPU load by adding enlightenments that enable the hypervisor synthetic interrupt controller (SynIC) and stimer. For example, with this enhancement, the host CPU load of a virtual machine running an idle Windows 10 guest should be approximately 0-5%.	Story Points:	---
Clone Of:	1610461	Environment:
Last Closed:	2019-02-13 07:43:29 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	Virt	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	1610461, 1631439
Bug Blocks:	1624786, 1638835, 1690641

Description Michal Skrivanek 2018-10-31 12:30:20 UTC

[see the original bug for more details]

cloning to oVirt, we may need to use the two backported enlightenments, probably add a separate enum to existing "os.windows_xp.devices.hyperv.enabled.value = true" key. We need a different set (those 2 new flags) for Windows 10 or we can probably send it to older windows too (needs to be tested) on new enough hosts - so possibly only in 4.3 cluster level. If it's per cluster level and it works for all Windows versions we may not need to touch the osinfo variable...


+++ This bug was initially created as a clone of Bug #1610461 +++

Description of problem:

High CPU load for Windows 10 (Update 1803) Guests when idle

Version-Release number of selected component (if applicable):

All 'qemu-kvm' and 'qemu-kvm-ev' (qemu-kvm-ev-2.10.0-21.el7_5.4.1)

How reproducible:

Install the last Windows10 on kvm virtual machine with one CPU.

Steps to Reproduce:
1. Install the last Windows10 on kvm virtual machine with one CPU.
2. Wait, idle time inside the Windows. (CPU load 0%)
3. See, in Linux virtual machine load 20-30%.

Actual results:
Linux virtual machine load 20-30%

Expected results:
Linux virtual machine load 0-5%


Additional info:
Need this possibility:
<hyperv>
    <synic state='on'/>
    <stimer state='on'/>
</hyperv>

It was disabled here:
https://bugzilla.redhat.com/show_bug.cgi?id=1336517

This is original post:
https://forum.proxmox.com/threads/high-cpu-load-for-windows-10-guests-when-idle.44531/

Comment 1 Vitaly Kuznetsov 2018-11-08 15:02:06 UTC

(In reply to Michal Skrivanek from comment #0)
> [see the original bug for more details]
> 
> cloning to oVirt, we may need to use the two backported enlightenments,
> probably add a separate enum to existing
> "os.windows_xp.devices.hyperv.enabled.value = true" key. We need a different
> set (those 2 new flags) for Windows 10 or we can probably send it to older
> windows too (needs to be tested)

This is my understanding too: all Windows versions should benefit from these enlightenments, we're not aware of any possible regressions ATM.

Comment 4 Michal Skrivanek 2018-12-03 09:49:46 UTC

this requires qemu-kvm-rhev-2.12.0-18.el7_6.2 and kernel newer than 3.10.0-957.1.3.el7. Not yet released

Comment 5 Michal Skrivanek 2018-12-03 09:50:18 UTC

*** Bug 1655039 has been marked as a duplicate of this bug. ***

Comment 6 Michal Skrivanek 2018-12-03 09:51:47 UTC

*** Bug 1655340 has been marked as a duplicate of this bug. ***

Comment 7 Michal Skrivanek 2018-12-03 09:52:43 UTC

WA is to use non-Windows OS type for now, or 4.2 cluster level

Comment 8 Liran Rotenberg 2019-01-23 14:23:43 UTC

Verified on:
ovirt-engine-4.3.0-0.8.master.20190120162615.git5926f20.el7.noarch
vdsm-4.30.6-17.gitcfd81b7.el7.x86_64
qemu-kvm-rhev-2.12.0-18.el7_6.2.x86_64
libvirt-daemon-kvm-4.5.0-10.el7_6.3.x86_64
qemu-kvm-tools-rhev-2.12.0-18.el7_6.2.x86_64
qemu-kvm-common-rhev-2.12.0-18.el7_6.2.x86_64
qemu-kvm-rhev-debuginfo-2.12.0-18.el7_6.2.x86_64
kernel-3.10.0-970.el7.x86_64

Steps:
1. Create a windows 10 VM, with update 1803, OS Type set to: Other OS.
2. Run the VM.
3. Check that the VM started without hyperv (virsh -r dumpxml <vm>) 
See it without:
<features>
    <acpi/>
    <hyperv>
      <relaxed state='on'/>
      <vapic state='on'/>
      <spinlocks state='on' retries='8191'/>
      <synic state='on'/>
      <stimer state='on'/>
    </hyperv>
</features>
4. Run on the host:
# top -p <vm_pid> -n 1800 -d 2 -b > unfixed-top-pc-result
5. Stop the VM.
6. Change the OS Type to Windows 10 64b.
7. Run the VM.
8. Check that the VM started with hyperv (virsh -r dumpxml <vm>)
See it has:
<features>
    <acpi/>
    <hyperv>
      <relaxed state='on'/>
      <vapic state='on'/>
      <spinlocks state='on' retries='8191'/>
      <synic state='on'/>
      <stimer state='on'/>
    </hyperv>
</features>
9. Run on the host:
# top -p <vm_pid> -n 1800 -d 2 -b > fixed-top-pc-result
10. Run on both outputs( unfixed-top-pc-result, fixed-top-pc-result)
# cat <output> |grep qemu-kvm|awk -F ' ' '{print $9;}'|awk '{sum+=$1} END {print "Average = ", sum/NR}'
11. Check the difference from step 10.

Results:
From unfixed-top-pc-result, the output of the command in step 11: Average =  46.3238
From fixed-top-pc-result, the output of the command in step 11: Average =  11.3415

Comment 9 Sandro Bonazzola 2019-01-24 12:18:29 UTC

*** Bug 1669102 has been marked as a duplicate of this bug. ***

Comment 10 Rolfe Dlugy-Hegwer 2019-02-08 02:01:21 UTC

Hi Steve,

Would you review the updated content in the Doc Text field? 

Thanks,

Rolfe

Comment 11 Steven Rosenberg 2019-02-10 09:54:40 UTC

Dear Rolfe, 

The doc text you added does basically fit the description of this issue.

With Best Regards.

Steven Rosenberg.

Comment 12 Sandro Bonazzola 2019-02-13 07:43:29 UTC

This bugzilla is included in oVirt 4.3.0 release, published on February 4th 2019.

Since the problem described in this bug report should be
resolved in oVirt 4.3.0 release, it has been closed with a resolution of CURRENT RELEASE.

If the solution does not work for you, please open a new bug report.