Bug 1880326

Summary:	[release 4.5] cluster-kube-scheduler-operator: Fix bug in reflector not recovering from "Too large resource version"
Product:	OpenShift Container Platform	Reporter:	Lukasz Szaszkiewicz <lszaszki>
Component:	kube-scheduler	Assignee:	Maciej Szulik <maszulik>
Status:	CLOSED ERRATA	QA Contact:	RamaKasturi <knarra>
Severity:	high	Docs Contact:
Priority:	high
Version:	4.5	CC:	aos-bugs, mfojtik, palonsor
Target Milestone:	---
Target Release:	4.5.z
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:	2020-11-17 16:06:00 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:
Bug Depends On:	1880369
Bug Blocks:	1879901

Description Lukasz Szaszkiewicz 2020-09-18 09:58:06 UTC

A recent fix in the reflector/informer https://github.com/kubernetes/kubernetes/pull/92688 prevents components/operators from entering a hotloop and stuck.

There are already reported cases that have run into that issue and were stuck for hours or even days. For example https://bugzilla.redhat.com/show_bug.cgi?id=1877346.

The root cause of the issue is the fact that a watch cache is initialized from the global revision (etcd) and might stay on it for an undefined period (if no changes were (add, modify) made).
That means that the watch cache across server instances may be out of sync.
That might lead to a situation in which a client gets a resource version from a server that has observed a newer rv, disconnect (due to a network error) from it, and reconnect to a server that is behind, resulting in “Too large resource version“ errors.

More details in https://github.com/kubernetes/kubernetes/issues/91073 and https://github.com/kubernetes/enhancements/tree/master/keps/sig-api-machinery/1904-efficient-watch-resumption

It looks like the issue only affects 1.18. According to https://github.com/kubernetes/kubernetes/issues/91073#issuecomment-652251669 the issue was first introduced in that version by changes done to the reflector.
The fix is already present in 1.19.

Please make sure that cluster-kube-scheduler-operator is using a client-go that includes https://github.com/kubernetes/kubernetes/pull/92688 if not please use this BZ and file a PR.
In case you are using a framework to build your operator make sure it uses the right version of the client-go library.

Comment 2 Maciej Szulik 2020-10-01 14:18:50 UTC

PR in the queue.

Comment 3 Maciej Szulik 2020-10-23 10:30:09 UTC

PR in the queue.

Comment 5 RamaKasturi 2020-11-09 09:52:53 UTC

Verified with payload below and i see that fix is present

4.5.0-0.nightly-2020-11-07-002125


[knarra@knarra cluster-kube-scheduler-operator]$ oc adm release info --commits registry.svc.ci.openshift.org/ocp/release:4.5.0-0.nightly-2020-11-07-002125 | grep cluster-kube-scheduler-operator
  cluster-kube-scheduler-operator                https://github.com/openshift/cluster-kube-scheduler-operator                a8218a72d631f93990c46f1337873954b2e66d88
[knarra@knarra cluster-kube-scheduler-operator]$ grep 'k8s.io/client-go ' go.mod
	k8s.io/client-go v0.18.9
[knarra@knarra cluster-kube-scheduler-operator]$ git log --date local --pretty="%h %an %cd - %s" a8218a7 | grep "#287"
c419f3bd OpenShift Merge Robot Thu Oct 15 07:09:40 2020 - Merge pull request #287 from soltysh/bug1880326

Based on the above moving bug to verified state.

Comment 8 errata-xmlrpc 2020-11-17 16:06:00 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 (OpenShift Container Platform 4.5.19 bug fix 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/RHBA-2020:5051

Comment 9 RamaKasturi 2021-08-25 07:26:22 UTC

Setting qe_test_coverage flag to '+' as i see that regression test coverage would be sufficient to test this bug.