123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257 |
- package weed_server
- import (
- "context"
- "crypto/md5"
- "hash"
- "io"
- "io/ioutil"
- "net/http"
- "os"
- "path"
- "strconv"
- "strings"
- "time"
- "github.com/chrislusf/seaweedfs/weed/filer2"
- "github.com/chrislusf/seaweedfs/weed/glog"
- "github.com/chrislusf/seaweedfs/weed/operation"
- "github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
- "github.com/chrislusf/seaweedfs/weed/security"
- "github.com/chrislusf/seaweedfs/weed/stats"
- "github.com/chrislusf/seaweedfs/weed/util"
- )
- func (fs *FilerServer) autoChunk(ctx context.Context, w http.ResponseWriter, r *http.Request,
- replication string, collection string, dataCenter string, ttlSec int32, ttlString string, fsync bool) {
- // autoChunking can be set at the command-line level or as a query param. Query param overrides command-line
- query := r.URL.Query()
- parsedMaxMB, _ := strconv.ParseInt(query.Get("maxMB"), 10, 32)
- maxMB := int32(parsedMaxMB)
- if maxMB <= 0 && fs.option.MaxMB > 0 {
- maxMB = int32(fs.option.MaxMB)
- }
- chunkSize := 1024 * 1024 * maxMB
- stats.FilerRequestCounter.WithLabelValues("postAutoChunk").Inc()
- start := time.Now()
- defer func() {
- stats.FilerRequestHistogram.WithLabelValues("postAutoChunk").Observe(time.Since(start).Seconds())
- }()
- var reply *FilerPostResult
- var err error
- var md5bytes []byte
- if r.Method == "POST" {
- reply, md5bytes, err = fs.doPostAutoChunk(ctx, w, r, chunkSize, replication, collection, dataCenter, ttlSec, ttlString, fsync)
- } else {
- reply, md5bytes, err = fs.doPutAutoChunk(ctx, w, r, chunkSize, replication, collection, dataCenter, ttlSec, ttlString, fsync)
- }
- if err != nil {
- writeJsonError(w, r, http.StatusInternalServerError, err)
- } else if reply != nil {
- if len(md5bytes) > 0 {
- w.Header().Set("Content-MD5", util.Base64Encode(md5bytes))
- }
- writeJsonQuiet(w, r, http.StatusCreated, reply)
- }
- }
- func (fs *FilerServer) doPostAutoChunk(ctx context.Context, w http.ResponseWriter, r *http.Request, chunkSize int32, replication string, collection string, dataCenter string, ttlSec int32, ttlString string, fsync bool) (filerResult *FilerPostResult, md5bytes []byte, replyerr error) {
- multipartReader, multipartReaderErr := r.MultipartReader()
- if multipartReaderErr != nil {
- return nil, nil, multipartReaderErr
- }
- part1, part1Err := multipartReader.NextPart()
- if part1Err != nil {
- return nil, nil, part1Err
- }
- fileName := part1.FileName()
- if fileName != "" {
- fileName = path.Base(fileName)
- }
- contentType := part1.Header.Get("Content-Type")
- if contentType == "application/octet-stream" {
- contentType = ""
- }
- fileChunks, md5Hash, chunkOffset, err := fs.uploadReaderToChunks(w, r, part1, chunkSize, replication, collection, dataCenter, ttlString, fileName, contentType, fsync)
- if err != nil {
- return nil, nil, err
- }
- fileChunks, replyerr = filer2.MaybeManifestize(fs.saveAsChunk(replication, collection, dataCenter, ttlString, fsync), fileChunks)
- if replyerr != nil {
- glog.V(0).Infof("manifestize %s: %v", r.RequestURI, replyerr)
- return
- }
- md5bytes = md5Hash.Sum(nil)
- filerResult, replyerr = fs.saveMetaData(ctx, r, fileName, replication, collection, ttlSec, contentType, md5bytes, fileChunks, chunkOffset)
- return
- }
- func (fs *FilerServer) doPutAutoChunk(ctx context.Context, w http.ResponseWriter, r *http.Request, chunkSize int32, replication string, collection string, dataCenter string, ttlSec int32, ttlString string, fsync bool) (filerResult *FilerPostResult, md5bytes []byte, replyerr error) {
- fileName := ""
- contentType := ""
- fileChunks, md5Hash, chunkOffset, err := fs.uploadReaderToChunks(w, r, r.Body, chunkSize, replication, collection, dataCenter, ttlString, fileName, contentType, fsync)
- if err != nil {
- return nil, nil, err
- }
- fileChunks, replyerr = filer2.MaybeManifestize(fs.saveAsChunk(replication, collection, dataCenter, ttlString, fsync), fileChunks)
- if replyerr != nil {
- glog.V(0).Infof("manifestize %s: %v", r.RequestURI, replyerr)
- return
- }
- md5bytes = md5Hash.Sum(nil)
- filerResult, replyerr = fs.saveMetaData(ctx, r, fileName, replication, collection, ttlSec, contentType, md5bytes, fileChunks, chunkOffset)
- return
- }
- func (fs *FilerServer) saveMetaData(ctx context.Context, r *http.Request, fileName string, replication string, collection string, ttlSec int32, contentType string, md5bytes []byte, fileChunks []*filer_pb.FileChunk, chunkOffset int64) (filerResult *FilerPostResult, replyerr error) {
- // detect file mode
- modeStr := r.URL.Query().Get("mode")
- if modeStr == "" {
- modeStr = "0660"
- }
- mode, err := strconv.ParseUint(modeStr, 8, 32)
- if err != nil {
- glog.Errorf("Invalid mode format: %s, use 0660 by default", modeStr)
- mode = 0660
- }
- // fix the path
- path := r.URL.Path
- if strings.HasSuffix(path, "/") {
- if fileName != "" {
- path += fileName
- }
- }
- // fix the crTime
- existingEntry, err := fs.filer.FindEntry(ctx, util.FullPath(path))
- crTime := time.Now()
- if err == nil && existingEntry != nil {
- crTime = existingEntry.Crtime
- }
- glog.V(4).Infoln("saving", path)
- entry := &filer2.Entry{
- FullPath: util.FullPath(path),
- Attr: filer2.Attr{
- Mtime: time.Now(),
- Crtime: crTime,
- Mode: os.FileMode(mode),
- Uid: OS_UID,
- Gid: OS_GID,
- Replication: replication,
- Collection: collection,
- TtlSec: ttlSec,
- Mime: contentType,
- Md5: md5bytes,
- },
- Chunks: fileChunks,
- }
- filerResult = &FilerPostResult{
- Name: fileName,
- Size: chunkOffset,
- }
- if dbErr := fs.filer.CreateEntry(ctx, entry, false, false); dbErr != nil {
- fs.filer.DeleteChunks(entry.Chunks)
- replyerr = dbErr
- filerResult.Error = dbErr.Error()
- glog.V(0).Infof("failing to write %s to filer server : %v", path, dbErr)
- }
- return filerResult, replyerr
- }
- func (fs *FilerServer) uploadReaderToChunks(w http.ResponseWriter, r *http.Request, reader io.Reader, chunkSize int32, replication string, collection string, dataCenter string, ttlString string, fileName string, contentType string, fsync bool) ([]*filer_pb.FileChunk, hash.Hash, int64, error) {
- var fileChunks []*filer_pb.FileChunk
- md5Hash := md5.New()
- var partReader = ioutil.NopCloser(io.TeeReader(reader, md5Hash))
- chunkOffset := int64(0)
- for {
- limitedReader := io.LimitReader(partReader, int64(chunkSize))
- // assign one file id for one chunk
- fileId, urlLocation, auth, assignErr := fs.assignNewFileInfo(replication, collection, dataCenter, ttlString, fsync)
- if assignErr != nil {
- return nil, nil, 0, assignErr
- }
- // upload the chunk to the volume server
- uploadResult, uploadErr := fs.doUpload(urlLocation, w, r, limitedReader, fileName, contentType, nil, auth)
- if uploadErr != nil {
- return nil, nil, 0, uploadErr
- }
- // if last chunk exhausted the reader exactly at the border
- if uploadResult.Size == 0 {
- break
- }
- // Save to chunk manifest structure
- fileChunks = append(fileChunks, uploadResult.ToPbFileChunk(fileId, chunkOffset))
- glog.V(4).Infof("uploaded %s chunk %d to %s [%d,%d)", fileName, len(fileChunks), fileId, chunkOffset, chunkOffset+int64(uploadResult.Size))
- // reset variables for the next chunk
- chunkOffset = chunkOffset + int64(uploadResult.Size)
- // if last chunk was not at full chunk size, but already exhausted the reader
- if int64(uploadResult.Size) < int64(chunkSize) {
- break
- }
- }
- return fileChunks, md5Hash, chunkOffset, nil
- }
- func (fs *FilerServer) doUpload(urlLocation string, w http.ResponseWriter, r *http.Request, limitedReader io.Reader, fileName string, contentType string, pairMap map[string]string, auth security.EncodedJwt) (*operation.UploadResult, error) {
- stats.FilerRequestCounter.WithLabelValues("postAutoChunkUpload").Inc()
- start := time.Now()
- defer func() {
- stats.FilerRequestHistogram.WithLabelValues("postAutoChunkUpload").Observe(time.Since(start).Seconds())
- }()
- uploadResult, err, _ := operation.Upload(urlLocation, fileName, fs.option.Cipher, limitedReader, false, contentType, pairMap, auth)
- return uploadResult, err
- }
- func (fs *FilerServer) saveAsChunk(replication string, collection string, dataCenter string, ttlString string, fsync bool) filer2.SaveDataAsChunkFunctionType {
- return func(reader io.Reader, name string, offset int64) (*filer_pb.FileChunk, string, string, error) {
- // assign one file id for one chunk
- fileId, urlLocation, auth, assignErr := fs.assignNewFileInfo(replication, collection, dataCenter, ttlString, fsync)
- if assignErr != nil {
- return nil, "", "", assignErr
- }
- // upload the chunk to the volume server
- uploadResult, uploadErr, _ := operation.Upload(urlLocation, name, fs.option.Cipher, reader, false, "", nil, auth)
- if uploadErr != nil {
- return nil, "", "", uploadErr
- }
- return uploadResult.ToPbFileChunk(fileId, offset), collection, replication, nil
- }
- }
|