Bug 1347979

Summary: CPU QoS does not work
Product: [oVirt] ovirt-engine Reporter: Artyom <alukiano>
Component: Backend.CoreAssignee: Nobody <nobody>
Status: CLOSED DUPLICATE QA Contact: Aharon Canan <acanan>
Severity: urgent Docs Contact:
Priority: unspecified    
Version: 4.0.0CC: alukiano, bugs, dfediuck, ncredi
Target Milestone: ---Keywords: Automation, Regression
Target Release: ---Flags: gklein: ovirt-4.0.0?
gklein: blocker?
rule-engine: planning_ack?
rule-engine: devel_ack?
rule-engine: testing_ack?
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2016-06-20 08:26:01 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: SLA RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
host vdsm and libvirtd log
none
new vdsm and mom logs none

Description Artyom 2016-06-19 13:55:50 UTC 
Created attachment 1169530 [details]
host vdsm and libvirtd log

Description of problem:
CPU QoS does not work

Version-Release number of selected component (if applicable):
rhevm-4.0.0.4-0.1.el7ev.noarch
vdsm-4.18.3-0.el7ev.x86_64
libvirt-1.2.17-13.el7_2.5.x86_64

How reproducible:
Always

Steps to Reproduce:
1. Create CPU QoS on datacenter with the limit 10
2. Create   profile with above QoS
3. Create VM with number of CPU equal to number of CPU's on the host_1 and with above CPU profile
4. Start VM on the host_1
5. Load VM CPU to 100%

Actual results:
Host CPU also has load 100%

Expected results:
Host CPU has load 10%

Additional info:
I can see via virsh that we send correct quota
<name>golden_env_mixed_virtio_0</name>
  <uuid>cbe701c0-f395-4b90-a67f-a461057f4638</uuid>
  <metadata xmlns:ovirt="http://ovirt.org/vm/tune/1.0">
    <ovirt:qos xmlns:ovirt="http://ovirt.org/vm/tune/1.0">
        <ovirt:vcpuLimit>10</ovirt:vcpuLimit>
</ovirt:qos>
  </metadata>
but from some reason values under cgroups are not updated:
# cat /sys/fs/cgroup/cpu\,cpuacct/machine.slice/machine-qemu\\x2d128\\x2dgoldenenvmixedvirtio0.scope/vcpu0/cpu.cfs_period_us 
100000
# cat /sys/fs/cgroup/cpu\,cpuacct/machine.slice/machine-qemu\\x2d128\\x2dgoldenenvmixedvirtio0.scope/vcpu0/cpu.cfs_quota_us 
-1
Comment 1 Artyom 2016-06-19 14:18:09 UTC 
I can say I have some similar problem connect to CPU shares.
I choose CPU shares equal to 512 and under cgroups I have:
[root@master-vds10 machine-qemu\x2d179\x2dgoldenenvmixedvirtio0.scope]# cat cpu.shares
512
[root@master-vds10 machine-qemu\x2d179\x2dgoldenenvmixedvirtio0.scope]# cd vcpu0/
[root@master-vds10 vcpu0]# cat cpu.shares
1024

Looks like CPU shares value applied only on VM but does not apply on VCPU
Comment 2 Artyom 2016-06-20 07:43:12 UTC 
I believe it can be connected to the bug https://bugzilla.redhat.com/show_bug.cgi?id=1346252
Comment 3 Martin Sivák 2016-06-20 08:02:00 UTC 
Artyom, can you please attach mom.log (with DEBUG evel enabled)? I believe you will see something like Guest xyz starting and Guest xyz is missing data.

MOM version would be nice too.
Comment 4 Doron Fediuck 2016-06-20 08:26:01 UTC 
Bug 1346252 had the same root cause of applying the policy.
Let's focus on a single BZ for the same issue.

*** This bug has been marked as a duplicate of bug 1346252 ***
Comment 5 Artyom 2016-06-20 08:55:04 UTC 
Created attachment 1169793 [details]
new vdsm and mom logs
Comment 6 Artyom 2016-06-20 08:56:20 UTC 
Mom version: mom-0.5.4-1.el7ev.noarch