Add support of fallocate for FS and XL backends (#3032)

cmd/fallocate.go

@@ -0,0 +1,25 @@
+// +build !linux
+
+/*
+ * Minio Cloud Storage, (C) 2016 Minio, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package cmd
+
+// Fallocate is not POSIX and not supported under Windows
+// Always return successful
+func Fallocate(fd int, offset int64, len int64) error {
+	return nil
+}

cmd/fallocate_linux.go

@@ -0,0 +1,31 @@
+// +build linux
+
+/*
+ * Minio Cloud Storage, (C) 2016 Minio, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package cmd
+
+import "syscall"
+
+// Fallocate uses the linux Fallocate syscall, which helps us to be
+// sure that subsequent writes on a file just created will not fail,
+// in addition, file allocation will be contigous on the disk
+func Fallocate(fd int, offset int64, len int64) error {
+	return syscall.Fallocate(fd,
+		1, // FALLOC_FL_KEEP_SIZE
+		offset,
+		len)
+}

cmd/fs-v1-multipart.go

@@ -450,6 +450,15 @@ func (fs fsObjects) PutObjectPart(bucket, object, uploadID string, partID int, s
 		bufSize = size
 	}
 	buf := make([]byte, int(bufSize))
+
+	if size > 0 {
+		// Prepare file to avoid disk fragmentation
+		err := fs.storage.PrepareFile(minioMetaBucket, tmpPartPath, size)
+		if err != nil {
+			return "", toObjectErr(err, minioMetaBucket, tmpPartPath)
+		}
+	}
+
 	bytesWritten, cErr := fsCreateFile(fs.storage, teeReader, buf, minioMetaBucket, tmpPartPath)
 	if cErr != nil {
 		fs.storage.DeleteFile(minioMetaBucket, tmpPartPath)
@@ -599,6 +608,18 @@ func (fs fsObjects) ListObjectParts(bucket, object, uploadID string, partNumberM
 	return fs.listObjectParts(bucket, object, uploadID, partNumberMarker, maxParts)
 }
 
+func (fs fsObjects) totalObjectSize(fsMeta fsMetaV1, parts []completePart) (int64, error) {
+	objSize := int64(0)
+	for _, part := range parts {
+		partIdx := fsMeta.ObjectPartIndex(part.PartNumber)
+		if partIdx == -1 {
+			return 0, InvalidPart{}
+		}
+		objSize += fsMeta.Parts[partIdx].Size
+	}
+	return objSize, nil
+}
+
 // CompleteMultipartUpload - completes an ongoing multipart
 // transaction after receiving all the parts indicated by the client.
 // Returns an md5sum calculated by concatenating all the individual
@@ -668,6 +689,19 @@ func (fs fsObjects) CompleteMultipartUpload(bucket string, object string, upload
 
 		// Allocate staging buffer.
 		var buf = make([]byte, readSizeV1)
+		var objSize int64
+
+		objSize, err = fs.totalObjectSize(fsMeta, parts)
+		if err != nil {
+			return "", traceError(err)
+		}
+		if objSize > 0 {
+			// Prepare file to avoid disk fragmentation
+			err = fs.storage.PrepareFile(minioMetaBucket, tempObj, objSize)
+			if err != nil {
+				return "", traceError(err)
+			}
+		}
 
 		// Loop through all parts, validate them and then commit to disk.
 		for i, part := range parts {

cmd/fs-v1-multipart_test.go

@@ -90,13 +90,18 @@ func TestPutObjectPartFaultyDisk(t *testing.T) {
 	for i := 1; i <= 7; i++ {
 		// Faulty disk generates errFaultyDisk at 'i' storage api call number
 		fs.storage = newNaughtyDisk(fsStorage, map[int]error{i: errFaultyDisk}, nil)
-		if _, err := fs.PutObjectPart(bucketName, objectName, uploadID, 1, dataLen, bytes.NewReader(data), md5Hex, sha256sum); errorCause(err) != errFaultyDisk {
+		md5sum, err := fs.PutObjectPart(bucketName, objectName, uploadID, 1, dataLen, bytes.NewReader(data), md5Hex, sha256sum)
+		if errorCause(err) != errFaultyDisk {
+			if errorCause(err) == nil {
+				t.Fatalf("Test %d shouldn't succeed, md5sum = %s\n", i, md5sum)
+			}
 			switch i {
 			case 1:
 				if !isSameType(errorCause(err), BucketNotFound{}) {
 					t.Fatal("Unexpected error ", err)
 				}
-			case 2, 4:
+			case 3:
+			case 2, 4, 5:
 				if !isSameType(errorCause(err), InvalidUploadID{}) {
 					t.Fatal("Unexpected error ", err)
 				}

cmd/fs-v1.go

@@ -388,6 +388,15 @@ func (fs fsObjects) PutObject(bucket string, object string, size int64, data io.
 			return ObjectInfo{}, toObjectErr(traceError(err), bucket, object)
 		}
 	} else {
+
+		// Prepare file to avoid disk fragmentation
+		if size > 0 {
+			err = fs.storage.PrepareFile(minioMetaBucket, tempObj, size)
+			if err != nil {
+				return ObjectInfo{}, toObjectErr(err, bucket, object)
+			}
+		}
+
 		// Allocate a buffer to Read() from request body
 		bufSize := int64(readSizeV1)
 		if size > 0 && bufSize > size {

cmd/naughty-disk_test.go

@@ -108,6 +108,13 @@ func (d *naughtyDisk) ReadFile(volume string, path string, offset int64, buf []b
 	return d.disk.ReadFile(volume, path, offset, buf)
 }
 
+func (d *naughtyDisk) PrepareFile(volume, path string, length int64) error {
+	if err := d.calcError(); err != nil {
+		return err
+	}
+	return d.disk.PrepareFile(volume, path, length)
+}
+
 func (d *naughtyDisk) AppendFile(volume, path string, buf []byte) error {
 	if err := d.calcError(); err != nil {
 		return err

cmd/posix-errors.go

@@ -22,6 +22,26 @@ import (
 	"syscall"
 )
 
+// Function not implemented error
+func isSysErrNoSys(err error) bool {
+	return err != nil && err == syscall.ENOSYS
+}
+
+// Not supported error
+func isSysErrOpNotSupported(err error) bool {
+	return err != nil && err == syscall.EOPNOTSUPP
+}
+
+// No space left on device error
+func isSysErrNoSpace(err error) bool {
+	return err != nil && err == syscall.ENOSPC
+}
+
+// Input/output error
+func isSysErrIO(err error) bool {
+	return err != nil && err == syscall.EIO
+}
+
 // Check if the given error corresponds to ENOTDIR (is not a directory)
 func isSysErrNotDir(err error) bool {
 	if pathErr, ok := err.(*os.PathError); ok {

cmd/posix.go

@@ -546,6 +546,121 @@ func (s *posix) ReadFile(volume string, path string, offset int64, buf []byte) (
 	return int64(m), err
 }
 
+func (s *posix) createFile(volume, path string) (f *os.File, err error) {
+	defer func() {
+		if err == syscall.EIO {
+			atomic.AddInt32(&s.ioErrCount, 1)
+		}
+	}()
+
+	if s.ioErrCount > maxAllowedIOError {
+		return nil, errFaultyDisk
+	}
+
+	// Validate if disk is free.
+	if err = s.checkDiskFree(); err != nil {
+		return nil, err
+	}
+
+	volumeDir, err := s.getVolDir(volume)
+	if err != nil {
+		return nil, err
+	}
+	// Stat a volume entry.
+	_, err = os.Stat(preparePath(volumeDir))
+	if err != nil {
+		if os.IsNotExist(err) {
+			return nil, errVolumeNotFound
+		}
+		return nil, err
+	}
+
+	filePath := pathJoin(volumeDir, path)
+	if err = checkPathLength(filePath); err != nil {
+		return nil, err
+	}
+
+	// Verify if the file already exists and is not of regular type.
+	var st os.FileInfo
+	if st, err = os.Stat(preparePath(filePath)); err == nil {
+		if !st.Mode().IsRegular() {
+			return nil, errIsNotRegular
+		}
+	} else {
+		// Create top level directories if they don't exist.
+		// with mode 0777 mkdir honors system umask.
+		if err = mkdirAll(preparePath(slashpath.Dir(filePath)), 0777); err != nil {
+			// File path cannot be verified since one of the parents is a file.
+			if isSysErrNotDir(err) {
+				return nil, errFileAccessDenied
+			} else if isSysErrPathNotFound(err) {
+				// Add specific case for windows.
+				return nil, errFileAccessDenied
+			}
+			return nil, err
+		}
+	}
+
+	w, err := os.OpenFile(preparePath(filePath), os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0666)
+	if err != nil {
+		// File path cannot be verified since one of the parents is a file.
+		if isSysErrNotDir(err) {
+			return nil, errFileAccessDenied
+		}
+		return nil, err
+	}
+
+	return w, nil
+}
+
+// PrepareFile - run prior actions before creating a new file for optimization purposes
+// Currenty we use fallocate when available to avoid disk fragmentation as much as possible
+func (s *posix) PrepareFile(volume, path string, fileSize int64) (err error) {
+
+	// It doesn't make sense to create a negative-sized file
+	if fileSize <= 0 {
+		return errInvalidArgument
+	}
+
+	defer func() {
+		if err == syscall.EIO {
+			atomic.AddInt32(&s.ioErrCount, 1)
+		}
+	}()
+
+	if s.ioErrCount > maxAllowedIOError {
+		return errFaultyDisk
+	}
+
+	// Create file if not found
+	w, err := s.createFile(volume, path)
+	if err != nil {
+		return err
+	}
+
+	// Close upon return.
+	defer w.Close()
+
+	// Allocate needed disk space to append data
+	e := Fallocate(int(w.Fd()), 0, fileSize)
+
+	// Ignore errors when Fallocate is not supported in the current system
+	if e != nil && !isSysErrNoSys(e) && !isSysErrOpNotSupported(e) {
+		switch {
+		case isSysErrNoSpace(e):
+			err = errDiskFull
+		case isSysErrIO(e):
+			err = e
+		default:
+			// For errors: EBADF, EINTR, EINVAL, ENODEV, EPERM, ESPIPE  and ETXTBSY
+			// Appending was failed anyway, returns unexpected error
+			err = errUnexpected
+		}
+		return err
+	}
+	return nil
+}
+
 // AppendFile - append a byte array at path, if file doesn't exist at
 // path this call explicitly creates it.
 func (s *posix) AppendFile(volume, path string, buf []byte) (err error) {
@@ -559,57 +674,11 @@ func (s *posix) AppendFile(volume, path string, buf []byte) (err error) {
 		return errFaultyDisk
 	}
 
-	// Validate if disk is free.
+	// Create file if not found
-	if err = s.checkDiskFree(); err != nil {
+	w, err := s.createFile(volume, path)
-		return err
-	}
-
-	volumeDir, err := s.getVolDir(volume)
 	if err != nil {
 		return err
 	}
-	// Stat a volume entry.
-	_, err = os.Stat(preparePath(volumeDir))
-	if err != nil {
-		if os.IsNotExist(err) {
-			return errVolumeNotFound
-		}
-		return err
-	}
-	filePath := pathJoin(volumeDir, path)
-	if err = checkPathLength(filePath); err != nil {
-		return err
-	}
-	// Verify if the file already exists and is not of regular type.
-	var st os.FileInfo
-	if st, err = os.Stat(preparePath(filePath)); err == nil {
-		if !st.Mode().IsRegular() {
-			return errIsNotRegular
-		}
-	}
-	// Create top level directories if they don't exist.
-	// with mode 0777 mkdir honors system umask.
-	if err = mkdirAll(preparePath(slashpath.Dir(filePath)), 0777); err != nil {
-		// File path cannot be verified since one of the parents is a file.
-		if isSysErrNotDir(err) {
-			return errFileAccessDenied
-		} else if isSysErrPathNotFound(err) {
-			// Add specific case for windows.
-			return errFileAccessDenied
-		}
-		return err
-	}
-
-	// Creates the named file with mode 0666 (before umask), or starts appending
-	// to an existig file.
-	w, err := os.OpenFile(preparePath(filePath), os.O_CREATE|os.O_APPEND|os.O_WRONLY, 0666)
-	if err != nil {
-		// File path cannot be verified since one of the parents is a file.
-		if isSysErrNotDir(err) {
-			return errFileAccessDenied
-		}
-		return err
-	}
 
 	// Close upon return.
 	defer w.Close()

cmd/posix_test.go

@@ -1126,6 +1126,100 @@ func TestAppendFile(t *testing.T) {
 	}
 }
 
+// Test posix.PrepareFile()
+func TestPrepareFile(t *testing.T) {
+	// create posix test setup
+	posixStorage, path, err := newPosixTestSetup()
+	if err != nil {
+		t.Fatalf("Unable to create posix test setup, %s", err)
+	}
+	defer removeAll(path)
+
+	// Setup test environment.
+	if err = posixStorage.MakeVol("success-vol"); err != nil {
+		t.Fatalf("Unable to create volume, %s", err)
+	}
+
+	if err = os.Mkdir(slashpath.Join(path, "success-vol", "object-as-dir"), 0777); err != nil {
+		t.Fatalf("Unable to create directory, %s", err)
+	}
+
+	testCases := []struct {
+		fileName    string
+		expectedErr error
+	}{
+		{"myobject", nil},
+		{"path/to/my/object", nil},
+		// Test to append to previously created file.
+		{"myobject", nil},
+		// Test to use same path of previously created file.
+		{"path/to/my/testobject", nil},
+		{"object-as-dir", errIsNotRegular},
+		// path segment uses previously uploaded object.
+		{"myobject/testobject", errFileAccessDenied},
+		// One path segment length is > 255 chars long.
+		{"path/to/my/object0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001", errFileNameTooLong},
+	}
+
+	// Add path length > 1024 test specially as OS X system does not support 1024 long path.
+	err = errFileNameTooLong
+	if runtime.GOOS != "darwin" {
+		err = nil
+	}
+	// path length is 1024 chars long.
+	testCases = append(testCases, struct {
+		fileName    string
+		expectedErr error
+	}{"level0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001/level0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000002/level0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000003/object000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001", err})
+
+	for _, testCase := range testCases {
+		if err = posixStorage.PrepareFile("success-vol", testCase.fileName, 16); err != testCase.expectedErr {
+			t.Errorf("Case: %s, expected: %s, got: %s", testCase, testCase.expectedErr, err)
+		}
+	}
+
+	// Test for permission denied.
+	if runtime.GOOS != "windows" {
+		// Initialize posix storage layer for permission denied error.
+		posixStorage, err = newPosix("/usr")
+		if err != nil {
+			t.Fatalf("Unable to initialize posix, %s", err)
+		}
+
+		if err = posixStorage.PrepareFile("bin", "yes", 16); !os.IsPermission(err) {
+			t.Errorf("expected: Permission error, got: %s", err)
+		}
+	}
+
+	// Test case with invalid file size which should be strictly positive
+	err = posixStorage.PrepareFile("bn", "yes", -3)
+	if err != errInvalidArgument {
+		t.Fatalf("should fail: %v", err)
+	}
+
+	// Test case with invalid volume name.
+	// A valid volume name should be atleast of size 3.
+	err = posixStorage.PrepareFile("bn", "yes", 16)
+	if err != errInvalidArgument {
+		t.Fatalf("expected: \"Invalid argument error\", got: \"%s\"", err)
+	}
+
+	// Test case with IO error count > max limit.
+
+	// setting ioErrCnt to 6.
+	// should fail with errFaultyDisk.
+	if posixType, ok := posixStorage.(*posix); ok {
+		// setting the io error count from as specified in the test case.
+		posixType.ioErrCount = int32(6)
+		err = posixType.PrepareFile("abc", "yes", 16)
+		if err != errFaultyDisk {
+			t.Fatalf("Expected \"Faulty Disk\", got: \"%s\"", err)
+		}
+	} else {
+		t.Fatalf("Expected the StorageAPI to be of type *posix")
+	}
+}
+
 // Test posix.RenameFile()
 func TestRenameFile(t *testing.T) {
 	// create posix test setup

cmd/storage-interface.go

@@ -35,6 +35,7 @@ type StorageAPI interface {
 	// File operations.
 	ListDir(volume, dirPath string) ([]string, error)
 	ReadFile(volume string, path string, offset int64, buf []byte) (n int64, err error)
+	PrepareFile(volume string, path string, len int64) (err error)
 	AppendFile(volume string, path string, buf []byte) (err error)
 	RenameFile(srcVolume, srcPath, dstVolume, dstPath string) error
 	StatFile(volume string, path string) (file FileInfo, err error)

cmd/storage-rpc-client.go

@@ -186,6 +186,18 @@ func (n networkStorage) DeleteVol(volume string) error {
 
 // File operations.
 
+func (n networkStorage) PrepareFile(volume, path string, length int64) (err error) {
+	reply := GenericReply{}
+	if err = n.rpcClient.Call("Storage.PrepareFileHandler", &PrepareFileArgs{
+		Vol:  volume,
+		Path: path,
+		Size: length,
+	}, &reply); err != nil {
+		return toStorageErr(err)
+	}
+	return nil
+}
+
 // CreateFile - create file.
 func (n networkStorage) AppendFile(volume, path string, buffer []byte) (err error) {
 	reply := GenericReply{}

cmd/storage-rpc-client_test.go

@@ -243,6 +243,10 @@ func (s *TestRPCStorageSuite) testRPCStorageFileOps(t *testing.T) {
 		if err != nil {
 			t.Error("Unable to initiate MakeVol", err)
 		}
+		err = storageDisk.PrepareFile("myvol", "file1", int64(len([]byte("Hello, world"))))
+		if err != nil {
+			t.Error("Unable to initiate AppendFile", err)
+		}
 		err = storageDisk.AppendFile("myvol", "file1", []byte("Hello, world"))
 		if err != nil {
 			t.Error("Unable to initiate AppendFile", err)

cmd/storage-rpc-server-datatypes.go

@@ -61,6 +61,21 @@ type ReadFileArgs struct {
 	Size int
 }
 
+// PrepareFileArgs represents append file RPC arguments.
+type PrepareFileArgs struct {
+	// Authentication token generated by Login.
+	GenericArgs
+
+	// Name of the volume.
+	Vol string
+
+	// Name of the path.
+	Path string
+
+	// Size of the file to be prepared
+	Size int64
+}
+
 // AppendFileArgs represents append file RPC arguments.
 type AppendFileArgs struct {
 	// Authentication token generated by Login.

cmd/storage-rpc-server.go

@@ -180,6 +180,14 @@ func (s *storageServer) ReadFileHandler(args *ReadFileArgs, reply *[]byte) (err
 	return err
 }
 
+// PrepareFileHandler - prepare file handler is rpc wrapper to prepare file.
+func (s *storageServer) PrepareFileHandler(args *PrepareFileArgs, reply *GenericReply) error {
+	if !isRPCTokenValid(args.Token) {
+		return errInvalidToken
+	}
+	return s.storage.PrepareFile(args.Vol, args.Path, args.Size)
+}
+
 // AppendFileHandler - append file handler is rpc wrapper to append file.
 func (s *storageServer) AppendFileHandler(args *AppendFileArgs, reply *GenericReply) error {
 	if !isRPCTokenValid(args.Token) {

cmd/xl-v1-common.go

@@ -69,3 +69,16 @@ func (xl xlObjects) isObject(bucket, prefix string) (ok bool) {
 	} // Exhausted all disks - return false.
 	return false
 }
+
+// Calculate the space occupied by an object in a single disk
+func (xl xlObjects) sizeOnDisk(fileSize int64, blockSize int64, dataBlocks int) int64 {
+	numBlocks := fileSize / blockSize
+	chunkSize := getChunkSize(blockSize, dataBlocks)
+	sizeInDisk := numBlocks * chunkSize
+	remaining := fileSize % blockSize
+	if remaining > 0 {
+		sizeInDisk += getChunkSize(remaining, dataBlocks)
+	}
+
+	return sizeInDisk
+}

cmd/xl-v1-multipart.go

@@ -424,6 +424,15 @@ func (xl xlObjects) PutObjectPart(bucket, object, uploadID string, partID int, s
 	// Delete the temporary object part. If PutObjectPart succeeds there would be nothing to delete.
 	defer xl.deleteObject(minioMetaBucket, tmpPartPath)
 
+	if size > 0 {
+		for _, disk := range onlineDisks {
+			if disk != nil {
+				actualSize := xl.sizeOnDisk(size, xlMeta.Erasure.BlockSize, xlMeta.Erasure.DataBlocks)
+				disk.PrepareFile(minioMetaBucket, tmpPartPath, actualSize)
+			}
+		}
+	}
+
 	// Erasure code data and write across all disks.
 	sizeWritten, checkSums, err := erasureCreateFile(onlineDisks, minioMetaBucket, tmpPartPath, teeReader, xlMeta.Erasure.BlockSize, xl.dataBlocks, xl.parityBlocks, bitRotAlgo, xl.writeQuorum)
 	if err != nil {

cmd/xl-v1-object.go

@@ -448,6 +448,15 @@ func (xl xlObjects) PutObject(bucket string, object string, size int64, data io.
 	// delete.
 	defer xl.deleteObject(minioMetaTmpBucket, tempObj)
 
+	if size > 0 {
+		for _, disk := range onlineDisks {
+			if disk != nil {
+				actualSize := xl.sizeOnDisk(size, xlMeta.Erasure.BlockSize, xlMeta.Erasure.DataBlocks)
+				disk.PrepareFile(minioMetaBucket, tempErasureObj, actualSize)
+			}
+		}
+	}
+
 	// Erasure code data and write across all disks.
 	sizeWritten, checkSums, err := erasureCreateFile(onlineDisks, minioMetaBucket, tempErasureObj, teeReader, xlMeta.Erasure.BlockSize, xlMeta.Erasure.DataBlocks, xlMeta.Erasure.ParityBlocks, bitRotAlgo, xl.writeQuorum)
 	if err != nil {