Bug 1949238

Summary:	Warning about clearing node selector / node name appears on every migration plan
Product:	Migration Toolkit for Containers	Reporter:	Derek Whatley <dwhatley>
Component:	Controller	Assignee:	Derek Whatley <dwhatley>
Status:	CLOSED ERRATA	QA Contact:	Xin jiang <xjiang>
Severity:	unspecified	Docs Contact:	Avital Pinnick <apinnick>
Priority:	unspecified
Version:	1.4.3	CC:	akarol, ernelson, rjohnson, sregidor, whu, xjiang
Target Milestone:	---
Target Release:	1.4.3
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2021-04-28 05:52:41 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Derek Whatley 2021-04-13 18:39:41 UTC

Description of problem:
There is a warning about NodeSelector/NodeName getting cleared upon migration, originally implemented in https://github.com/konveyor/mig-controller/pull/946, that triggers on every Migration Plan due to default NodeName/NodeSelector practices on certain versions of OCP:

- NodeName appears to always get set on OCP 3.11 (triggers the warning)
- NodeSelector appears to automatically get set in some cases to ["node-role.kubernetes.io/compute": "true",] (triggers the warning)

We think that it's inappropriate to warn the user in these conditions since this is completely expected and there is no action needed on the users behalf. The user DOES need to perhaps intervene if they've configured custom nodeselectors. We should still raise the warning if a non-standard nodeselector is assigned.

Version-Release number of selected component (if applicable):
Exists in 1.4.2 release, 1.4.3 pre-release images. Plan to fix prior to 1.4.3 release.


PR to show this warning more selectively is posted here https://github.com/konveyor/mig-controller/pull/1069.

NEEDINFO from QE on the default nodeselector and nodename practices across the spectrum of OCP versions (3.7 - 4.x)

Comment 1 Derek Whatley 2021-04-13 18:41:02 UTC

Oh, it was also reported that this warning disappears after a migration is started. This should also be fixed.

Comment 2 Aziza Karol 2021-04-19 08:57:10 UTC

Is this targeted to be fixed in 1.4.3 or should this be moved to 1.4.4?

Comment 3 Derek Whatley 2021-04-19 14:13:47 UTC

Waiting on info from QE on default nodeselectors across cluster versions. I have a PR ready but just need to add any additional defaults that we shouldn't warn on.

Comment 4 Xin jiang 2021-04-20 05:42:13 UTC

we just tried OCP 3.9/3.10/3.11, by default NodeSelector appears to automatically set label ["node-role.kubernetes.io/compute": "true",]  on pod. Because of PSI issue, still unable to get OCP 3.7 cluster

Comment 8 Xin jiang 2021-04-21 08:44:39 UTC

verified with MTC 1.4.3.

registry.redhat.io/rhmtc/openshift-migration-controller-rhel8@sha256:bc34f77b2df5b5da18c46b77066fbb239ea4a2ff4517fc280d21bd87909fd69a

Comment 13 errata-xmlrpc 2021-04-28 05:52:41 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 (Migration Toolkit for Containers (MTC) tool image release advisory 1.4.3), 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/RHBA-2021:1439