Bug 1849169

Summary:	[RFE] add virtualCPUs/physicalCPUs ratio property to evenly_distributed policy
Product:	Red Hat Enterprise Virtualization Manager	Reporter:	Andrea Perotti <aperotti>
Component:	ovirt-engine	Assignee:	Lucia Jelinkova <ljelinko>
Status:	CLOSED ERRATA	QA Contact:	Tadeas Kozub <tkozub>
Severity:	medium	Docs Contact:
Priority:	unspecified
Version:	4.4.0	CC:	ahadas, gdeolive, ljelinko, mavital, michal.skrivanek, schandle
Target Milestone:	ovirt-4.5.0	Keywords:	FutureFeature, ZStream
Target Release:	4.5.0
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:	ovirt-engine-4.5.0	Doc Type:	Enhancement
Doc Text:	Feature: A new parameter was added to the evenly_distributed scheduling policy that takes into account the ratio between virtual and physical CPUs on the host. Reason: To prevent the host from over utilization of all physical CPUs. Result: When the ratio is set to 0, the evenly distributed policy works as before. If the value is greater than 0, the vCPU to physical CPU is considered as follows: a. when scheduling a VM, hosts with lower CPU utilization are preferred. However, if adding of the VM would cause the vCPU to physical ratio to be exceeded, the hosts vCPU to physical ratio AND cpu utilization are considered. b. in a running environment, if the host's vCPU to physical ratio is above the limit, some of the VMs might be load balanced to the hosts with lower vCPU to physical CPU ratio.	Story Points:	---
Clone Of:
Clones:	1981101 (view as bug list)		Environment:
Last Closed:	2022-05-26 16:22:27 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:
Bug Blocks:	1789090, 1981101

Description Andrea Perotti 2020-06-19 18:18:39 UTC

1. What is the nature and description of the request?  

A new property is requested for evenly_distributed policy, in addition to the current ones, that takes in account also the virtualCPUs/physicalCPUs ratio


, to produce automatic VMs live migration in case of a ratio greater than a proper value (possible values: decimal number in the range 0.1 - 2.5) This would allow to avoid CPU overcommit and rebalance the loads.

2. Why would you need this? (List the business requirements here)

During RHV cluster upgrade or other workload rebalance situation, is important for CPU intensive workloads to avoid CPU overcommit on each hypervisor.

3. How would you like to achieve this? (List the functional requirements here)  

Add a new property in the evenly_distributed policy named "virtualCPUs/physicalCPUs ratio" 
to accept decimal number values in the range 0.1 - 2.5


4. For each functional requirement listed, specify how you can test to confirm the requirement is successfully implemented.   

- Create cluster with 2 hypervisor
- the 2 nodes (A and B) have 1 dual core CPU per each
- Set the "virtualCPUs/physicalCPUs ratio" property in the evenly_distributed policy to 1.5
- Create new 4 VMs and start all of them on node A
- Scheduler should start only 3 VMs on A, 4th should start on node B

5. Do you have any specific timeline dependencies ?  
As soon as possible

6. Would you be able to assist in testing this functionality if implemented?
yes

Comment 4 Arik 2021-12-06 08:09:50 UTC

Lucia, can you please add documentation for this?

Comment 7 Tadeas Kozub 2022-04-28 12:13:31 UTC

Verified on: Software Version:4.5.0.4-0.1.el8ev

Tested with 2 hosts in a cluster, each host has 32 CPUs and each VM 12 vCPUs. Using 4 VMs.
When cluster policy set to "evenly_distributed", VCpuToPhysicalCpuRatio set to 1.25 and Specific Host set 
all 4 VMs start on this specific host, shortly after, one of the VMs migrates to the other host as intended.

Comment 8 Tadeas Kozub 2022-04-28 12:20:40 UTC

Apr 28, 2022, 2:17:21 PM
Migration completed (VM: vm1, Source: host1, Destination: host2, Duration: 6 seconds, Total: 6 seconds, Actual downtime: (N/A))

Apr 28, 2022, 2:17:14 PM
Migration initiated by system (VM: vm1, Source: host1, Destination: host2, Reason: Load balancing).

Apr 28, 2022, 2:17:03 PM
VM vm2 started on Host host1

Apr 28, 2022, 2:17:03 PM
VM vm1 started on Host host1

Apr 28, 2022, 2:17:03 PM
VM vm4 started on Host host1

Apr 28, 2022, 2:17:03 PM
VM vm3 started on Host host1

Comment 13 errata-xmlrpc 2022-05-26 16:22:27 UTC

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (Moderate: RHV Manager (ovirt-engine) [ovirt-4.5.0] security update), and where to find the updated
files, follow the link below.

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

https://access.redhat.com/errata/RHSA-2022:4711

Comment 14 meital avital 2022-08-08 13:47:43 UTC

Due to QE capacity, we are not going to cover this issue in our automation