Prefer local disks when fetching data blocks (#9563)

If the requested server is part of the set this will always read 
from the local disk, even if the disk contains a parity shard. 
In default setup there is a 50% chance that at least 
one shard that otherwise would have been fetched remotely 
will be read locally instead.

It basically trades RPC call overhead for reed-solomon. 
On distributed localhost this seems to be fairly break-even, 
with a very small gain in throughput and latency. 
However on networked servers this should be a bigger

1MB objects, before:

```
Operation: GET. Concurrency: 32. Hosts: 4.

Requests considered: 76257:
 * Avg: 25ms 50%: 24ms 90%: 32ms 99%: 42ms Fastest: 7ms Slowest: 67ms
 * First Byte: Average: 23ms, Median: 22ms, Best: 5ms, Worst: 65ms

Throughput:
* Average: 1213.68 MiB/s, 1272.63 obj/s (59.948s, starting 14:45:44 CEST)
```

After:
```
Operation: GET. Concurrency: 32. Hosts: 4.

Requests considered: 78845:
 * Avg: 24ms 50%: 24ms 90%: 31ms 99%: 39ms Fastest: 8ms Slowest: 62ms
 * First Byte: Average: 22ms, Median: 21ms, Best: 6ms, Worst: 57ms

Throughput:
* Average: 1255.11 MiB/s, 1316.08 obj/s (59.938s, starting 14:43:58 CEST)
```

Bonus fix: Only ask for heal once on an object.

cmd/erasure-decode.go

@@ -31,22 +31,59 @@ var errHealRequired = errors.New("heal required")
 // Reads in parallel from readers.
 type parallelReader struct {
 	readers       []io.ReaderAt
+	orgReaders    []io.ReaderAt
 	dataBlocks    int
 	offset        int64
 	shardSize     int64
 	shardFileSize int64
 	buf           [][]byte
+	readerToBuf   []int
 }
 
 // newParallelReader returns parallelReader.
 func newParallelReader(readers []io.ReaderAt, e Erasure, offset, totalLength int64) *parallelReader {
+	r2b := make([]int, len(readers))
+	for i := range r2b {
+		r2b[i] = i
+	}
 	return &parallelReader{
-		readers,
+		readers:       readers,
-		e.dataBlocks,
+		orgReaders:    readers,
-		(offset / e.blockSize) * e.ShardSize(),
+		dataBlocks:    e.dataBlocks,
-		e.ShardSize(),
+		offset:        (offset / e.blockSize) * e.ShardSize(),
-		e.ShardFileSize(totalLength),
+		shardSize:     e.ShardSize(),
-		make([][]byte, len(readers)),
+		shardFileSize: e.ShardFileSize(totalLength),
+		buf:           make([][]byte, len(readers)),
+		readerToBuf:   r2b,
+	}
+}
+
+// preferReaders can mark readers as preferred.
+// These will be chosen before others.
+func (p *parallelReader) preferReaders(prefer []bool) {
+	if len(prefer) != len(p.orgReaders) {
+		return
+	}
+	// Copy so we don't change our input.
+	tmp := make([]io.ReaderAt, len(p.orgReaders))
+	copy(tmp, p.orgReaders)
+	p.readers = tmp
+	// next is the next non-preferred index.
+	next := 0
+	for i, ok := range prefer {
+		if !ok || p.readers[i] == nil {
+			continue
+		}
+		if i == next {
+			next++
+			continue
+		}
+		// Move reader with index i to index next.
+		// Do this by swapping next and i
+		p.readers[next], p.readers[i] = p.readers[i], p.readers[next]
+		p.readerToBuf[next] = i
+		p.readerToBuf[i] = next
+		next++
 	}
 }
 
@@ -54,7 +91,7 @@ func newParallelReader(readers []io.ReaderAt, e Erasure, offset, totalLength int
 func (p *parallelReader) canDecode(buf [][]byte) bool {
 	bufCount := 0
 	for _, b := range buf {
-		if b != nil {
+		if len(b) > 0 {
 			bufCount++
 		}
 	}
@@ -62,13 +99,23 @@ func (p *parallelReader) canDecode(buf [][]byte) bool {
 }
 
 // Read reads from readers in parallel. Returns p.dataBlocks number of bufs.
-func (p *parallelReader) Read() ([][]byte, error) {
+func (p *parallelReader) Read(dst [][]byte) ([][]byte, error) {
-	newBuf := make([][]byte, len(p.readers))
+	newBuf := dst
+	if len(dst) != len(p.readers) {
+		newBuf = make([][]byte, len(p.readers))
+	} else {
+		for i := range newBuf {
+			newBuf[i] = newBuf[i][:0]
+		}
+	}
 	var newBufLK sync.RWMutex
 
 	if p.offset+p.shardSize > p.shardFileSize {
 		p.shardSize = p.shardFileSize - p.offset
 	}
+	if p.shardSize == 0 {
+		return newBuf, nil
+	}
 
 	readTriggerCh := make(chan bool, len(p.readers))
 	for i := 0; i < p.dataBlocks; i++ {
@@ -104,26 +151,30 @@ func (p *parallelReader) Read() ([][]byte, error) {
 				readTriggerCh <- true
 				return
 			}
-			if p.buf[i] == nil {
+			bufIdx := p.readerToBuf[i]
+			if p.buf[bufIdx] == nil {
 				// Reading first time on this disk, hence the buffer needs to be allocated.
 				// Subsequent reads will re-use this buffer.
-				p.buf[i] = make([]byte, p.shardSize)
+				p.buf[bufIdx] = make([]byte, p.shardSize)
 			}
 			// For the last shard, the shardsize might be less than previous shard sizes.
 			// Hence the following statement ensures that the buffer size is reset to the right size.
-			p.buf[i] = p.buf[i][:p.shardSize]
+			p.buf[bufIdx] = p.buf[bufIdx][:p.shardSize]
-			_, err := disk.ReadAt(p.buf[i], p.offset)
+			_, err := disk.ReadAt(p.buf[bufIdx], p.offset)
 			if err != nil {
 				if _, ok := err.(*errHashMismatch); ok {
 					atomic.StoreInt32(&healRequired, 1)
 				}
+
+				// This will be communicated upstream.
+				p.orgReaders[bufIdx] = nil
 				p.readers[i] = nil
 				// Since ReadAt returned error, trigger another read.
 				readTriggerCh <- true
 				return
 			}
 			newBufLK.Lock()
-			newBuf[i] = p.buf[i]
+			newBuf[bufIdx] = p.buf[bufIdx]
 			newBufLK.Unlock()
 			// Since ReadAt returned success, there is no need to trigger another read.
 			readTriggerCh <- false
@@ -156,8 +207,9 @@ func (err *errDecodeHealRequired) Unwrap() error {
 }
 
 // Decode reads from readers, reconstructs data if needed and writes the data to the writer.
-func (e Erasure) Decode(ctx context.Context, writer io.Writer, readers []io.ReaderAt, offset, length, totalLength int64) error {
+// A set of preferred drives can be supplied. In that case they will be used and the data reconstructed.
-	healRequired, err := e.decode(ctx, writer, readers, offset, length, totalLength)
+func (e Erasure) Decode(ctx context.Context, writer io.Writer, readers []io.ReaderAt, offset, length, totalLength int64, prefer []bool) error {
+	healRequired, err := e.decode(ctx, writer, readers, offset, length, totalLength, prefer)
 	if healRequired {
 		return &errDecodeHealRequired{err}
 	}
@@ -166,7 +218,7 @@ func (e Erasure) Decode(ctx context.Context, writer io.Writer, readers []io.Read
 }
 
 // Decode reads from readers, reconstructs data if needed and writes the data to the writer.
-func (e Erasure) decode(ctx context.Context, writer io.Writer, readers []io.ReaderAt, offset, length, totalLength int64) (bool, error) {
+func (e Erasure) decode(ctx context.Context, writer io.Writer, readers []io.ReaderAt, offset, length, totalLength int64, prefer []bool) (bool, error) {
 	if offset < 0 || length < 0 {
 		logger.LogIf(ctx, errInvalidArgument)
 		return false, errInvalidArgument
@@ -180,12 +232,16 @@ func (e Erasure) decode(ctx context.Context, writer io.Writer, readers []io.Read
 	}
 
 	reader := newParallelReader(readers, e, offset, totalLength)
+	if len(prefer) == len(readers) {
+		reader.preferReaders(prefer)
+	}
 
 	startBlock := offset / e.blockSize
 	endBlock := (offset + length) / e.blockSize
 
 	var healRequired bool
 	var bytesWritten int64
+	var bufs [][]byte
 	for block := startBlock; block <= endBlock; block++ {
 		var blockOffset, blockLength int64
 		switch {
@@ -205,9 +261,11 @@ func (e Erasure) decode(ctx context.Context, writer io.Writer, readers []io.Read
 		if blockLength == 0 {
 			break
 		}
-		bufs, err := reader.Read()
+		var err error
+		bufs, err = reader.Read(bufs)
 		if err != nil {
 			if errors.Is(err, errHealRequired) {
+				// errHealRequired is only returned if there are be enough data for reconstruction.
 				healRequired = true
 			} else {
 				return healRequired, err

cmd/erasure-decode_test.go

@@ -138,7 +138,7 @@ func TestErasureDecode(t *testing.T) {
 		}
 
 		writer := bytes.NewBuffer(nil)
-		err = erasure.Decode(context.Background(), writer, bitrotReaders, test.offset, test.length, test.data)
+		err = erasure.Decode(context.Background(), writer, bitrotReaders, test.offset, test.length, test.data, nil)
 		closeBitrotReaders(bitrotReaders)
 		if err != nil && !test.shouldFail {
 			t.Errorf("Test %d: should pass but failed with: %v", i, err)
@@ -181,7 +181,7 @@ func TestErasureDecode(t *testing.T) {
 				bitrotReaders[0] = nil
 			}
 			writer.Reset()
-			err = erasure.Decode(context.Background(), writer, bitrotReaders, test.offset, test.length, test.data)
+			err = erasure.Decode(context.Background(), writer, bitrotReaders, test.offset, test.length, test.data, nil)
 			closeBitrotReaders(bitrotReaders)
 			if err != nil && !test.shouldFailQuorum {
 				t.Errorf("Test %d: should pass but failed with: %v", i, err)
@@ -191,7 +191,7 @@ func TestErasureDecode(t *testing.T) {
 			}
 			if !test.shouldFailQuorum {
 				if content := writer.Bytes(); !bytes.Equal(content, data[test.offset:test.offset+test.length]) {
-					t.Errorf("Test %d: read retruns wrong file content", i)
+					t.Errorf("Test %d: read returns wrong file content", i)
 				}
 			}
 		}
@@ -271,7 +271,7 @@ func TestErasureDecodeRandomOffsetLength(t *testing.T) {
 			tillOffset := erasure.ShardFileTillOffset(offset, readLen, length)
 			bitrotReaders[index] = newStreamingBitrotReader(disk, "testbucket", "object", tillOffset, DefaultBitrotAlgorithm, erasure.ShardSize())
 		}
-		err = erasure.Decode(context.Background(), buf, bitrotReaders, offset, readLen, length)
+		err = erasure.Decode(context.Background(), buf, bitrotReaders, offset, readLen, length, nil)
 		closeBitrotReaders(bitrotReaders)
 		if err != nil {
 			t.Fatal(err, offset, readLen)
@@ -333,7 +333,7 @@ func benchmarkErasureDecode(data, parity, dataDown, parityDown int, size int64,
 			tillOffset := erasure.ShardFileTillOffset(0, size, size)
 			bitrotReaders[index] = newStreamingBitrotReader(disk, "testbucket", "object", tillOffset, DefaultBitrotAlgorithm, erasure.ShardSize())
 		}
-		if err = erasure.Decode(context.Background(), bytes.NewBuffer(content[:0]), bitrotReaders, 0, size, size); err != nil {
+		if err = erasure.Decode(context.Background(), bytes.NewBuffer(content[:0]), bitrotReaders, 0, size, size, nil); err != nil {
 			panic(err)
 		}
 		closeBitrotReaders(bitrotReaders)

cmd/erasure-heal.go

@@ -28,7 +28,7 @@ import (
 func (e Erasure) Heal(ctx context.Context, readers []io.ReaderAt, writers []io.Writer, size int64) error {
 	r, w := io.Pipe()
 	go func() {
-		if err := e.Decode(ctx, w, readers, 0, size, size); err != nil {
+		if err := e.Decode(ctx, w, readers, 0, size, size, nil); err != nil {
 			w.CloseWithError(err)
 			return
 		}

cmd/erasure.go

@@ -87,14 +87,15 @@ func (e *Erasure) EncodeData(ctx context.Context, data []byte) ([][]byte, error)
 // It only decodes the data blocks but does not verify them.
 // It returns an error if the decoding failed.
 func (e *Erasure) DecodeDataBlocks(data [][]byte) error {
-	needsReconstruction := false
+	var isZero = 0
-	for _, b := range data[:e.dataBlocks] {
+	for _, b := range data[:] {
-		if b == nil {
+		if len(b) == 0 {
-			needsReconstruction = true
+			isZero++
 			break
 		}
 	}
-	if !needsReconstruction {
+	if isZero == 0 || isZero == len(data) {
+		// If all are zero, payload is 0 bytes.
 		return nil
 	}
 	return e.encoder().ReconstructData(data)

cmd/xl-v1-object.go

@@ -265,6 +265,7 @@ func (xl xlObjects) getObject(ctx context.Context, bucket, object string, startO
 		return toObjectErr(err, bucket, object)
 	}
 
+	var healOnce sync.Once
 	for ; partIndex <= lastPartIndex; partIndex++ {
 		if length == totalBytesRead {
 			break
@@ -284,6 +285,7 @@ func (xl xlObjects) getObject(ctx context.Context, bucket, object string, startO
 		tillOffset := erasure.ShardFileTillOffset(partOffset, partLength, partSize)
 		// Get the checksums of the current part.
 		readers := make([]io.ReaderAt, len(onlineDisks))
+		prefer := make([]bool, len(onlineDisks))
 		for index, disk := range onlineDisks {
 			if disk == OfflineDisk {
 				continue
@@ -292,14 +294,19 @@ func (xl xlObjects) getObject(ctx context.Context, bucket, object string, startO
 			partPath := pathJoin(object, fmt.Sprintf("part.%d", partNumber))
 			readers[index] = newBitrotReader(disk, bucket, partPath, tillOffset,
 				checksumInfo.Algorithm, checksumInfo.Hash, erasure.ShardSize())
+
+			// Prefer local disks
+			prefer[index] = disk.Hostname() == ""
 		}
-		err := erasure.Decode(ctx, writer, readers, partOffset, partLength, partSize)
+		err := erasure.Decode(ctx, writer, readers, partOffset, partLength, partSize, prefer)
 		// Note: we should not be defer'ing the following closeBitrotReaders() call as we are inside a for loop i.e if we use defer, we would accumulate a lot of open files by the time
 		// we return from this function.
 		closeBitrotReaders(readers)
 		if err != nil {
 			if decodeHealErr, ok := err.(*errDecodeHealRequired); ok {
-				go deepHealObject(pathJoin(bucket, object))
+				healOnce.Do(func() {
+					go deepHealObject(pathJoin(bucket, object))
+				})
 				err = decodeHealErr.err
 			}
 			if err != nil {