From d4ec8f26e8ac622693d0b21dbd672fbb5c73cc2f Mon Sep 17 00:00:00 2001
From: Michal Pryc <mpryc@redhat.com>
Date: Tue, 20 Aug 2024 09:54:16 +0200
Subject: [PATCH] Fix #8132: managed fields failure during Restore

This commit addresses issue #8132, where an error randomly
appears in the logs during the restore operation.

The error occurs due to a race condition when attempting
to patch managed fields on an object that has been modified
in the cluster. The error message indicates that the operation
cannot be fulfilled because the object has been modified,
suggesting that changes should be applied to the latest version.

To resolve this, a retry mechanism has been implemented in the restore
process when encountering this error, ensuring that managed fields
are properly restored without the error message appearing in the logs.

Signed-off-by: Michal Pryc <mpryc@redhat.com>
---
 pkg/restore/restore.go | 53 +++++++++++++++++++++++++++++-------------
 1 file changed, 37 insertions(+), 16 deletions(-)

diff --git a/pkg/restore/restore.go b/pkg/restore/restore.go
index 63c26538bac..331aa164f3a 100644
--- a/pkg/restore/restore.go
+++ b/pkg/restore/restore.go
@@ -47,6 +47,7 @@ import (
 	"k8s.io/client-go/dynamic/dynamicinformer"
 	corev1 "k8s.io/client-go/kubernetes/typed/core/v1"
 	"k8s.io/client-go/tools/cache"
+	"k8s.io/client-go/util/retry"
 	crclient "sigs.k8s.io/controller-runtime/pkg/client"
 
 	"github.com/vmware-tanzu/velero/internal/credentials"
@@ -1669,24 +1670,44 @@ func (ctx *restoreContext) restoreItem(obj *unstructured.Unstructured, groupReso
 	}
 
 	// restore the managedFields
-	withoutManagedFields := createdObj.DeepCopy()
-	createdObj.SetManagedFields(obj.GetManagedFields())
-	patchBytes, err := generatePatch(withoutManagedFields, createdObj)
-	if err != nil {
-		ctx.log.Errorf("error generating patch for managed fields %s: %v", kube.NamespaceAndName(obj), err)
-		errs.Add(namespace, err)
-		return warnings, errs, itemExists
-	}
-	if patchBytes != nil {
-		if _, err = resourceClient.Patch(name, patchBytes); err != nil {
-			ctx.log.Errorf("error patch for managed fields %s: %v", kube.NamespaceAndName(obj), err)
-			if !apierrors.IsNotFound(err) {
-				errs.Add(namespace, err)
-				return warnings, errs, itemExists
+	var firstPatchTry bool = true
+
+	err = retry.RetryOnConflict(retry.DefaultRetry, func() error {
+		var withoutManagedFields *unstructured.Unstructured
+		var patchBytes []byte
+		var patchErr error
+
+		if !firstPatchTry {
+			createdObj, patchErr = resourceClient.Get(name, metav1.GetOptions{})
+			if patchErr != nil {
+				ctx.log.Errorf("error fetching latest object for %s: %v", kube.NamespaceAndName(obj), patchErr)
+				return patchErr
 			}
-		} else {
-			ctx.log.Infof("the managed fields for %s is patched", kube.NamespaceAndName(obj))
 		}
+
+		firstPatchTry = false
+		withoutManagedFields = createdObj.DeepCopy()
+		createdObj.SetManagedFields(obj.GetManagedFields())
+		patchBytes, patchErr = generatePatch(withoutManagedFields, createdObj)
+		if patchErr != nil {
+			ctx.log.Errorf("error generating patch for managed fields %s: %v", kube.NamespaceAndName(obj), patchErr)
+			return patchErr
+		}
+		if patchBytes != nil {
+			_, patchErr = resourceClient.Patch(name, patchBytes)
+			if patchErr != nil {
+				ctx.log.Errorf("error patching managed fields %s: %v", kube.NamespaceAndName(obj), patchErr)
+				return patchErr
+			}
+			ctx.log.Infof("managed fields for %s patched successfully", kube.NamespaceAndName(obj))
+		}
+
+		return nil
+	})
+
+	if err != nil && !apierrors.IsNotFound(err) {
+		errs.Add(namespace, err)
+		return warnings, errs, itemExists
 	}
 
 	if groupResource == kuberesource.Pods {