[release-4.18] OCPBUGS-46474:catalog-operator: Delete Pods that were evicted (#3459)

This change adds another reason why a Pod could be detected as "dead",
namely when it was evicted by the kubelet. This can happen when there is
resource pressure on the Node. Then the reason will be
"TerminationByKubelet".

This addresses the issue described in
https://issues.redhat.com/browse/OCPBUGS-45490

Signed-off-by: Simon Krenger <[email protected]>
Upstream-repository: operator-lifecycle-manager
Upstream-commit: 8e39847016ba04a4d12599a8d101f4af77c389b5

Author: simonkrenger

staging/operator-lifecycle-manager/pkg/controller/registry/reconciler/grpc.go

@@ -531,6 +531,7 @@ func imageChanged(logger *logrus.Entry, updatePod *corev1.Pod, servingPods []*co
 func isPodDead(pod *corev1.Pod) bool {
 	for _, check := range []func(*corev1.Pod) bool{
 		isPodDeletedByTaintManager,
+		isPodTerminatedByKubelet,
 	} {
 		if check(pod) {
 			return true
@@ -551,6 +552,19 @@ func isPodDeletedByTaintManager(pod *corev1.Pod) bool {
 	return false
 }
 
+// This reason is set when the Pod was evicted due to resource pressure on the Node
+func isPodTerminatedByKubelet(pod *corev1.Pod) bool {
+	if pod.DeletionTimestamp == nil {
+		return false
+	}
+	for _, condition := range pod.Status.Conditions {
+		if condition.Type == corev1.DisruptionTarget && condition.Reason == "TerminationByKubelet" && condition.Status == corev1.ConditionTrue {
+			return true
+		}
+	}
+	return false
+}
+
 // imageID returns the ImageID of the primary catalog source container or an empty string if the image ID isn't available yet.
 // Note: the pod must be running and the container in a ready status to return a valid ImageID.
 func imageID(pod *corev1.Pod) string {