Bug 2223464 - Wrong cpuset.mems cgroup is set when setting numa tuning with "restrictive" mode.
Summary: Wrong cpuset.mems cgroup is set when setting numa tuning with "restrictive" m...
Keywords:
Status: VERIFIED
Alias: None
Product: Red Hat Enterprise Linux 8
Classification: Red Hat
Component: libvirt
Version: 8.9
Hardware: Unspecified
OS: Unspecified
unspecified
medium
Target Milestone: rc
: ---
Assignee: Michal Privoznik
QA Contact: liang cong
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2023-07-18 01:49 UTC by liang cong
Modified: 2023-08-07 07:54 UTC (History)
7 users (show)

Fixed In Version: libvirt-8.0.0-22.el8
Doc Type: Bug Fix
Doc Text:
Cause: For 'restrictive' numatune, libvirt wasn't setting cpuset.mems on domain startup, which kind of defeats the purpose of the configuration knob. Consequence: Domain was not restricted to run on configured NUMA nodes. Fix: Simple. Result: Libvirt now sets up cpuset CGroup controller for 'restrictive' numatune mode even on domain startup, so domain runs on desired NUMA nodes from the beginning.
Clone Of:
Environment:
Last Closed:
Type: Bug
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker RHELPLAN-162508 0 None None None 2023-07-18 01:49:50 UTC

Description liang cong 2023-07-18 01:49:04 UTC
Description of problem:Wrong cpuset.mems cgroup is set when setting numa tuning with "restrictive" mode.

Version-Release number of selected component (if applicable):
# rpm -q libvirt qemu-kvm
libvirt-8.0.0-21.module+el8.9.0+19166+e262ca96.x86_64
qemu-kvm-6.2.0-35.module+el8.9.0+19166+e262ca96.x86_64

How reproducible:
100%

Steps to Reproduce:
1 Start a guest with below setting:
<numatune>
          <memory mode="restrictive" nodeset="0"/>
  </numatune>

2 Check cpuset.mems cgroup setting
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d1\\x2dvm1.scope/libvirt/emulator/cpuset.mems
0-1
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d1\\x2dvm1.scope/libvirt/vcpu0/cpuset.mems
0-1
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d1\\x2dvm1.scope/libvirt/vcpu1/cpuset.mems
0-1

Actual results:
We could see that the cpuset.mems cgroup setting of the guest is different with numa tuning setting

Expected results:
With restrictive mode, the cpuset.mems cgroup setting should follow the numa tuning setting.

Comment 1 Michal Privoznik 2023-07-20 08:34:48 UTC
I think this is the same as RHEL-9 bug 2070380. Upstream it was fixed by the following commit:

commit 629282d8845407c1aff9a26f5dc026e15121f8cd
Author:     Michal Prívozník <mprivozn>
AuthorDate: Fri Apr 1 14:30:05 2022 +0200
Commit:     Michal Prívozník <mprivozn>
CommitDate: Thu Apr 7 12:12:11 2022 +0200

    lib: Set up cpuset controller for restrictive numatune

    The aim of 'restrictive' numatune mode is to rely solely on
    CGroups to have QEMU running on configured NUMA nodes.  However,
    we were never setting the cpuset controller when a domain was
    starting up. We are doing so only when
    virDomainSetNumaParameters() is called (aka live pinning).

    This is obviously wrong. Fortunately, fix is simple as
    'restrictive' is similar to 'strict' - every location where
    VIR_DOMAIN_NUMATUNE_MEM_STRICT occurs can be audited and
    VIR_DOMAIN_NUMATUNE_MEM_RESTRICTIVE case can be added.

    Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2070380
    Signed-off-by: Michal Privoznik <mprivozn>
    Reviewed-by: Ján Tomko <jtomko>

v8.3.0-rc1~138

Comment 3 liang cong 2023-07-25 02:47:29 UTC
Pre-verified on scratch build:
# rpm -q libvirt
libvirt-8.0.0-22.el8_rc.3ae0219671.x86_64

Test steps:
1. Define a guest with numatune configed xml:
<numatune>
   <memory mode="restrictive" nodeset="1"/>
</numatune>

2. Start the guest
# virsh start vm1
Domain 'vm1' started

3. Check the cpuset.mems cgroup setting
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/emulator/cpuset.mems
1
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/vcpu0/cpuset.mems
1
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/vcpu1/cpuset.mems
1

Also tested with "strict", "interleave", "preferred" modes and with guest numa node

Comment 7 liang cong 2023-08-07 07:54:18 UTC
Verified on build:
# rpm -q libvirt
libvirt-8.0.0-22.module+el8.9.0+19544+b3045133.x86_64

Test steps:
1. Define a guest with numatune configed xml:
<numatune>
   <memory mode="restrictive" nodeset="0"/>
</numatune>

2. Start the guest
# virsh start vm1
Domain 'vm1' started

3. Check the cpuset.mems cgroup setting
# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/emulator/cpuset.mems 
0

# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/vcpu0/cpuset.mems 
0

# cat /sys/fs/cgroup/cpuset/machine.slice/machine-qemu\\x2d4\\x2dvm1.scope/libvirt/vcpu1/cpuset.mems 
0

Also tested with "strict", "interleave", "preferred" modes and with guest numa node


Note You need to log in before you can comment on or make changes to this bug.