[release-1.6] Fix BZ-2080547: mis-read the priority class after update (

#2003)

* Fix a bug when updating the priority class

Since most of the priority class fiels are emutable, when updating it,
HCO remove and re-create it. Then HCO reads the new created priority
class in order to set its refertence in the HyperConverged
relatedObjects list.

The re-reading often fails with "no found" just because some timing and
cache issues.

This PR is fixing the issue by droping the re-read.

Signed-off-by: Nahshon Unna-Tsameret <nunnatsa@redhat.com>

* fix the functional test in kubevirtci

Signed-off-by: Nahshon Unna-Tsameret <nunnatsa@redhat.com>

Co-authored-by: Nahshon Unna-Tsameret <nunnatsa@redhat.com>

hack/check_update_priority_class.sh

@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+set -ex
+
+# edit priority class labels, causing HCO to recreate it, making sure that the relatedObjects list contains the new priorityClass
+
+# check that the new priority class was created and that it does not include the new label
+function check_priority_class() {
+  set -ex
+  OLD_PC_UID=$1
+
+  # make sure the object was actually changed
+  NEW_PC_UID=$(${KUBECTL_BINARY} get priorityclass -n ${INSTALLED_NAMESPACE} kubevirt-cluster-critical -o jsonpath='{.metadata.uid}')
+  [[ "${OLD_PC_UID}" != "${NEW_PC_UID}" ]]
+
+  # read the UID from the HyperConverged relatedObject list
+  NEW_PC_REF_UID=$(${KUBECTL_BINARY} get hco -n ${INSTALLED_NAMESPACE} kubevirt-hyperconverged -o json | jq -r '.status.relatedObjects[] | select(.kind == "PriorityClass" and .name == "kubevirt-cluster-critical") | .uid')
+
+  [[ "${NEW_PC_UID}" == "${NEW_PC_REF_UID}" ]]
+
+  # make sure the new label was removed:
+  TEST_LABEL=$(${KUBECTL_BINARY} get priorityclass -n ${INSTALLED_NAMESPACE} kubevirt-cluster-critical -o jsonpath='{.metadata.labels.test}')
+  [[ -z "${TEST_LABEL}" ]]
+}
+
+export -f check_priority_class
+
+OLD_PC_REF_UID=$(${KUBECTL_BINARY} get hco -n ${INSTALLED_NAMESPACE} kubevirt-hyperconverged -o json | jq -r '.status.relatedObjects[] | select(.kind == "PriorityClass" and .name == "kubevirt-cluster-critical") | .uid')
+OLD_PC_UID=$(${KUBECTL_BINARY} get priorityclass -n ${INSTALLED_NAMESPACE} kubevirt-cluster-critical -o jsonpath='{.metadata.uid}')
+[[ "${OLD_PC_REF_UID}" == "${OLD_PC_UID}" ]]
+
+${KUBECTL_BINARY} patch priorityclass kubevirt-cluster-critical -n ${INSTALLED_NAMESPACE} --type=json -p='[{"op": "add", "path": "/metadata/labels/test", "value": "test"}]'
+./hack/retry.sh 3 10 "check_priority_class ${OLD_PC_UID}"

hack/run-tests.sh

@@ -2,7 +2,7 @@
 
 set -euo pipefail
 
-INSTALLED_NAMESPACE=${INSTALLED_NAMESPACE:-"kubevirt-hyperconverged"}
+export INSTALLED_NAMESPACE=${INSTALLED_NAMESPACE:-"kubevirt-hyperconverged"}
 
 source hack/common.sh
 source cluster/kubevirtci.sh
@@ -55,5 +55,7 @@ ${KUBECTL_BINARY} label priorityclass kubevirt-cluster-critical app-
 sleep 10
 [[ $(${KUBECTL_BINARY} get priorityclass kubevirt-cluster-critical -o=jsonpath='{.metadata.labels.app}') == 'kubevirt-hyperconverged' ]]
 
+./hack/check_update_priority_class.sh
+
 # Check the webhook, to see if it allow deleteing of the HyperConverged CR
 ./hack/retry.sh 10 30 "${KUBECTL_BINARY} delete hco -n ${INSTALLED_NAMESPACE} kubevirt-hyperconverged"

pkg/controller/operands/kubevirt.go

@@ -13,7 +13,6 @@ import (
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
-	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/apimachinery/pkg/util/yaml"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 
@@ -580,11 +579,7 @@ func (h *kvPriorityClassHooks) updateCr(req *common.HcoRequest, Client client.Cl
 		return false, false, err
 	}
 
-	// update found object for object references
-	err = Client.Get(req.Ctx, types.NamespacedName{Name: found.Name, Namespace: found.Namespace}, found)
-	if err != nil {
-		return true, !req.HCOTriggered, err
-	}
+	pc.DeepCopyInto(found)
 
 	return true, !req.HCOTriggered, nil
 }