cli: adjust index file creation in upload-bin

In case of incomplete search result it will try to find each missed oid
and process it. In case of duplicates the first found will be in index
file.

Close #3647

Signed-off-by: Ekaterina Pavlova <ekt@morphbits.io>

cli/util/uploader.go

@@ -354,16 +354,28 @@ func uploadIndexFiles(ctx *cli.Context, p *pool.Pool, containerID cid.ID, accoun
 	}
 
 	var (
-		errCh                 = make(chan error)
 		buffer   = make([]byte, indexFileSize*oidSize)
-		oidCh                 = make(chan oid.ID, indexFileSize)
-		oidFetcherToProcessor = make(chan struct{}, indexFileSize)
-
+		doneCh   = make(chan struct{})
+		errCh    = make(chan error)
 		emptyOid = make([]byte, oidSize)
 	)
-	defer close(oidCh)
+
+	go func() {
+		defer close(doneCh)
+
+		// Main processing loop for each index file.
+		for i := existingIndexCount; i < expectedIndexCount; i++ {
+			// Start block parsing goroutines.
+			var (
+				// processedIndices is a mapping from position in buffer to the block index.
+				processedIndices sync.Map
+				wg               sync.WaitGroup
+				oidCh            = make(chan oid.ID, 2*maxParallelSearches)
+			)
+			wg.Add(maxParallelSearches)
 			for range maxParallelSearches {
 				go func() {
+					defer wg.Done()
 					for id := range oidCh {
 						var obj *object.Object
 						errRetr := retry(func() error {
@@ -386,67 +398,64 @@ func uploadIndexFiles(ctx *cli.Context, p *pool.Pool, containerID cid.ID, accoun
 							}
 							return
 						}
-				offset := (uint(blockIndex) % indexFileSize) * oidSize
-				id.Encode(buffer[offset:])
-				oidFetcherToProcessor <- struct{}{}
+						pos := uint(blockIndex) % indexFileSize
+						if _, ok := processedIndices.LoadOrStore(pos, blockIndex); !ok {
+							id.Encode(buffer[pos*oidSize:])
+						}
 					}
 				}()
 			}
 
-	for i := existingIndexCount; i < expectedIndexCount; i++ {
+			// Search for blocks within the index file range.
 			startIndex := i * indexFileSize
 			endIndex := startIndex + indexFileSize
-		go func() {
-			for j := int(startIndex); j < int(endIndex); j += searchBatchSize {
-				remaining := int(endIndex) - j
-				end := j + min(searchBatchSize, remaining)
+			objIDs := searchObjects(ctx.Context, p, containerID, account, blockAttributeKey, startIndex, endIndex, maxParallelSearches, errCh)
+			for id := range objIDs {
+				oidCh <- id
+			}
+			close(oidCh)
+			wg.Wait()
+			fmt.Fprintf(ctx.App.Writer, "Index file %d generated, checking for the missing blocks...\n", i)
 
-				prm = client.PrmObjectSearch{}
-				filters = object.NewSearchFilters()
-				filters.AddFilter(blockAttributeKey, fmt.Sprintf("%d", j), object.MatchNumGE)
-				filters.AddFilter(blockAttributeKey, fmt.Sprintf("%d", end), object.MatchNumLT)
-				prm.SetFilters(filters)
-				var objIDs []oid.ID
-				err := retry(func() error {
-					var errSearchIndex error
-					objIDs, errSearchIndex = neofs.ObjectSearch(ctx.Context, p, account.PrivateKey(), containerID.String(), prm)
-					return errSearchIndex
-				})
+			// Check if there are empty OIDs in the generated index file. This may happen
+			// if searchObjects has returned not all blocks within the requested range, ref.
+			// #3645. In this case, retry the search for every missing object.
+			var count int
 
-				if err != nil {
+			for idx := range indexFileSize {
+				if _, ok := processedIndices.Load(idx); !ok {
+					count++
+					objIDs = searchObjects(ctx.Context, p, containerID, account, blockAttributeKey, i*indexFileSize+idx, i*indexFileSize+idx+1, 1, errCh)
+					// Block object duplicates are allowed, we're OK with the first found result.
+					id, ok := <-objIDs
+					for range objIDs {
+					}
+					if !ok {
 						select {
-					case errCh <- fmt.Errorf("failed to search for objects from %d to %d for index file %d: %w", j, end, i, err):
+						case errCh <- fmt.Errorf("block %d is missing from the storage", i*indexFileSize+idx):
 						default:
 						}
 						return
 					}
+					processedIndices.Store(idx, id)
+					id.Encode(buffer[idx*oidSize:])
+				}
+			}
+			fmt.Fprintf(ctx.App.Writer, "%d missing block(s) processed for index file %d, uploading index file...\n", count, i)
 
-				for _, id := range objIDs {
-					oidCh <- id
-				}
-			}
-		}()
-
-		var completed int
-	waitLoop:
-		for {
-			select {
-			case err := <-errCh:
-				return err
-			case <-oidFetcherToProcessor:
-				completed++
-				if completed == int(indexFileSize) {
-					break waitLoop
-				}
-			}
-		}
 			// Check if there are empty OIDs in the generated index file. If it happens at
 			// this stage, then there's a bug in the code.
 			for k := 0; k < len(buffer); k += oidSize {
 				if slices.Compare(buffer[k:k+oidSize], emptyOid) == 0 {
-				return fmt.Errorf("empty OID found in index file %d at position %d (block index %d)", i, k/oidSize, i+uint(k/oidSize))
+					select {
+					case errCh <- fmt.Errorf("empty OID found in index file %d at position %d (block index %d)", i, k/oidSize, i*indexFileSize+uint(k/oidSize)):
+					default:
+					}
+					return
 				}
 			}
+
+			// Upload index file.
 			attrs := []object.Attribute{
 				*object.NewAttribute(attributeKey, strconv.Itoa(int(i))),
 				*object.NewAttribute("IndexSize", strconv.Itoa(int(indexFileSize))),
@@ -455,14 +464,83 @@ func uploadIndexFiles(ctx *cli.Context, p *pool.Pool, containerID cid.ID, accoun
 				return uploadObj(ctx.Context, p, signer, account.PrivateKey().GetScriptHash(), containerID, buffer, attrs, homomorphicHashingDisabled)
 			})
 			if err != nil {
-			return fmt.Errorf("failed to upload index file %d: %w", i, err)
+				select {
+				case errCh <- fmt.Errorf("failed to upload index file %d: %w", i, err):
+				default:
+				}
+				return
 			}
 			fmt.Fprintf(ctx.App.Writer, "Uploaded index file %d\n", i)
+			clear(buffer)
+		}
+	}()
+
+	select {
+	case err := <-errCh:
+		return err
+	case <-doneCh:
 	}
 
 	return nil
 }
 
+// searchObjects searches in parallel for objects with attribute GE startIndex and LT
+// endIndex. It returns a buffered channel of resulting object IDs and closes it once
+// OID search is finished. Errors are sent to errCh in a non-blocking way.
+func searchObjects(ctx context.Context, p *pool.Pool, containerID cid.ID, account *wallet.Account, blockAttributeKey string, startIndex, endIndex uint, maxParallelSearches int, errCh chan error) chan oid.ID {
+	var res = make(chan oid.ID, 2*searchBatchSize)
+	go func() {
+		var wg sync.WaitGroup
+		defer close(res)
+
+		for i := int(startIndex); i < int(endIndex); i += searchBatchSize * maxParallelSearches {
+			for j := range maxParallelSearches {
+				start := i + j*searchBatchSize
+				end := start + searchBatchSize
+
+				if start >= int(endIndex) {
+					break
+				}
+				if end > int(endIndex) {
+					end = int(endIndex)
+				}
+
+				wg.Add(1)
+				go func(start, end int) {
+					defer wg.Done()
+
+					prm := client.PrmObjectSearch{}
+					filters := object.NewSearchFilters()
+					filters.AddFilter(blockAttributeKey, fmt.Sprintf("%d", start), object.MatchNumGE)
+					filters.AddFilter(blockAttributeKey, fmt.Sprintf("%d", end), object.MatchNumLT)
+					prm.SetFilters(filters)
+
+					var objIDs []oid.ID
+					err := retry(func() error {
+						var errBlockSearch error
+						objIDs, errBlockSearch = neofs.ObjectSearch(ctx, p, account.PrivateKey(), containerID.String(), prm)
+						return errBlockSearch
+					})
+					if err != nil {
+						select {
+						case errCh <- fmt.Errorf("failed to search for block(s) from %d to %d: %w", start, end, err):
+						default:
+						}
+						return
+					}
+
+					for _, id := range objIDs {
+						res <- id
+					}
+				}(start, end)
+			}
+			wg.Wait()
+		}
+	}()
+
+	return res
+}
+
 // uploadObj uploads object to the container using provided settings.
 func uploadObj(ctx context.Context, p *pool.Pool, signer user.Signer, owner util.Uint160, containerID cid.ID, objData []byte, attrs []object.Attribute, homomorphicHashingDisabled bool) error {
 	var (