-
Bug
-
Resolution: Unresolved
-
Undefined
-
rhel-9.4, rhel-9.4.z
-
None
-
libvirt-10.9.0-1.el9
-
No
-
Moderate
-
ZStream
-
rhel-sst-virtualization
-
ssg_virtualization
-
18
-
2
-
Dev ack
-
False
-
-
None
-
None
-
Approved Blocker
-
Pass
-
Manual
-
-
10.9.0
-
-
None
What were you trying to do that didn't work?
In OpenShift Virtualization, I'm querying the metric 'kubevirt_vmi_vcpu_wait_seconds_total'. All values are 0.
When querying libvirt directly, I can confirm that the returned value is 0:
# virsh -r domstats centos9 | grep wait vcpu.0.wait=0 <-- vcpu.0.halt_wait_ns.sum=147337004989 vcpu.1.wait=0 <-- vcpu.1.halt_wait_ns.sum=163969618614
What is the impact of this issue to you?
Incorrect metrics for VMs.
Please provide the package NVR for which the bug is seen:
libvirt-10.0.0-6.7.el9_4.x86_64
qemu-kvm-8.2.0-11.el9_4.6.x86_64
kernel-5.14.0-427.37.1.el9_4.x86_64
How reproducible is this bug?:
100%
Steps to reproduce
- Using a RHEL 9.4 hypervisor, add the kernel cmdline option 'schedstats=enable'
- Run a VM and generate IO load.
- Check the vCPU wait stats: "virsh -r domstats VM_NAME |grep wait"
Expected results
IO wait correctly reported.
Actual results
IO wait is 0, however the kernel is reporting IO wait in the 'wait_sum' stat:
# cat /proc/cmdline BOOT_IMAGE=(hd0,gpt3)/vmlinuz-5.14.0-427.37.1.el9_4.x86_64 root=UUID=aec1c1e8-3576-4eb2-ab62-f62984e655a2 console=tty0 console=ttyS0,115200n8 no_timer_check net.ifnames=0 crashkernel=1G-4G:192M,4G-64G:256M,64G-:512M schedstats=enable # virsh -r domstats centos9 |grep wait vcpu.0.wait=0 vcpu.0.halt_wait_ns.sum=147337004989 vcpu.1.wait=0 vcpu.1.halt_wait_ns.sum=163969618614 # pstree -t -p 4506 qemu-kvm(4506)─┬─{CPU 0/KVM}(4518) ├─{CPU 1/KVM}(4519) ├─{IO mon_iothread}(4517) ├─{qemu-kvm}(4513) ├─{vnc_worker}(4521) ├─{worker}(4514) ├─{worker}(4573) ├─{worker}(4575) ├─{worker}(4576) ├─{worker}(4577) ├─{worker}(4578) ├─{worker}(4579) ├─{worker}(4580) ├─{worker}(4581) ├─{worker}(4582) └─{worker}(4583) # grep wait /proc/4506/task/{4518,4519}/sched /proc/4506/task/4518/sched:wait_start : 0.000000 /proc/4506/task/4518/sched:wait_max : 4.657302 /proc/4506/task/4518/sched:wait_sum : 951.535557 /proc/4506/task/4518/sched:wait_count : 162006 /proc/4506/task/4518/sched:iowait_sum : 0.000000 /proc/4506/task/4518/sched:iowait_count : 0 /proc/4506/task/4519/sched:wait_start : 0.000000 /proc/4506/task/4519/sched:wait_max : 3.558880 /proc/4506/task/4519/sched:wait_sum : 865.273765 /proc/4506/task/4519/sched:wait_count : 154549 /proc/4506/task/4519/sched:iowait_sum : 0.000000 /proc/4506/task/4519/sched:iowait_count : 0
- is blocked by
-
RHEL-50574 Rebase libvirt in RHEL-9.6.0
- Integration
- is depended on by
-
CNV-49442 kubevirt_vmi_vcpu_wait_seconds_total metric is always 0
- Closed
- is duplicated by
-
CNV-49442 kubevirt_vmi_vcpu_wait_seconds_total metric is always 0
- Closed
- links to
-
RHBA-2024:140248 libvirt bug fix and enhancement update