CachedResource replication reconciler should retry on AlreadyExists

If the CachedObject is missing, and Create fails with AlreadyExists,
the reconciler now returns an error and lets the work item to be requeued.
The previous code would, in that case, attempt to do an Update but without
setting the object's meta, the Update call would always fail (missing
resourceVersion and others...). Letting the item to be requeued is cleaner
and fixes object updates.

On-behalf-of: @SAP [email protected]
Signed-off-by: Robert Vasek <[email protected]>

Author: gman0

pkg/reconciler/cache/cachedresources/replication/replication_reconcile.go

@@ -295,17 +295,7 @@ func (r *replicationReconciler) reconcile(ctx context.Context, gvr schema.GroupV
 
 		logger.V(2).Info("Creating object in global cache")
 		_, err := r.createObject(ctx, clusterName, localCopy)
-		if err != nil {
-			if apierrors.IsAlreadyExists(err) {
-				// If the object already exists, try to update it instead
-				logger.V(2).Info("Object already exists in global cache, updating instead")
-				_, err = r.updateObject(ctx, clusterName, localCopy)
-			}
-			if err != nil {
-				return err
-			}
-		}
-		return nil
+		return err
 	}
 
 	// update global copy and compare