use renameAll instead of deleteAll for metacache-manager (#13005)

renameAll is cheaper, rely on background deletes instead.

cmd/erasure-multipart.go

@@ -122,7 +122,7 @@ func (er erasureObjects) renameAll(ctx context.Context, bucket, prefix string) {
 		wg.Add(1)
 		go func(disk StorageAPI) {
 			defer wg.Done()
-			disk.RenameFile(ctx, bucket, prefix, minioMetaTmpBucket, mustGetUUID())
+			disk.RenameFile(ctx, bucket, prefix, minioMetaTmpDeletedBucket, mustGetUUID())
 		}(disk)
 	}
 	wg.Wait()

cmd/erasure-server-pool.go

@@ -1421,20 +1421,8 @@ func (z *erasureServerPools) DeleteBucket(ctx context.Context, bucket string, fo
 	return nil
 }
 
-// deleteAll will delete a bucket+prefix unconditionally across all disks.
-// Note that set distribution is ignored so it should only be used in cases where
-// data is not distributed across sets.
-// Errors are logged but individual disk failures are not returned.
-func (z *erasureServerPools) deleteAll(ctx context.Context, bucket, prefix string) {
-	for _, servers := range z.serverPools {
-		for _, set := range servers.sets {
-			set.deleteAll(ctx, bucket, prefix)
-		}
-	}
-}
-
 // renameAll will rename bucket+prefix unconditionally across all disks to
-// minioMetaTmpBucket + unique uuid,
+// minioMetaTmpDeletedBucket + unique uuid,
 // Note that set distribution is ignored so it should only be used in cases where
 // data is not distributed across sets. Errors are logged but individual
 // disk failures are not returned.

cmd/metacache.go

@@ -149,5 +149,5 @@ func (m *metacache) delete(ctx context.Context) {
 		logger.LogIf(ctx, errors.New("metacache.delete: expected objAPI to be *erasureServerPools"))
 		return
 	}
-	ez.deleteAll(ctx, minioMetaBucket, metacachePrefixForID(m.bucket, m.id))
+	ez.renameAll(ctx, minioMetaBucket, metacachePrefixForID(m.bucket, m.id))
 }