Bug 1939922 - Descheduler profile triggers needless stateful pods needlessly
Summary: Descheduler profile triggers needless stateful pods needlessly
Keywords:
Status: CLOSED DEFERRED
Alias: None
Product: OpenShift Container Platform
Classification: Red Hat
Component: kube-scheduler
Version: 4.7
Hardware: Unspecified
OS: Unspecified
high
urgent
Target Milestone: ---
: 4.8.0
Assignee: Mike Dame
QA Contact: RamaKasturi
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2021-03-17 10:17 UTC by Fabian Deutsch
Modified: 2021-04-08 19:56 UTC (History)
5 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2021-04-08 19:56:44 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)

Description Fabian Deutsch 2021-03-17 10:17:51 UTC
Description of problem:
In CNV we are looking at leveraging the LowNodeUtilization strategy of the LifecycleAndUtilization profile to balance out a cluster after upgrade.

However, the PodLifeTime is also part of this profile. This will trigger needless VM live migrations every 24hrs, because running pods will be evicted by this strategy.

A profile is needed that is exclusively focused on balancing the cluster according to load - spreak retroactively, not proactively.

Version-Release number of selected component (if applicable):
4.7

How reproducible:
Always

Steps to Reproduce:
1. Create a CNV VM, wait 2 days
2. Run descheduler with LifecycleAndUtilization profile
3. 

Actual results:
VM will be rescheduled even if the cluster is balanced

Expected results:
VM will not be scheduled because the cluster is balanced

Additional info:

Comment 2 Mike Dame 2021-03-18 14:41:34 UTC
I commented in https://issues.redhat.com/browse/CNV-10289, but this will likely actually need an RFE to expose any additional settings in the Descheduler Operator

Comment 3 Gaurav Singh 2021-03-18 14:53:14 UTC
There RFE to track this request 

https://issues.redhat.com/browse/RFE-1711


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