Files
seaweedfs/weed/admin/handlers/file_browser_grpc.go
T
孙超 c80664ec21 s3: propagate storage rule fsync to volume server uploads (#10906)
The storage rule's fsync decision was computed by the filer
(detectStorageOption -> rule.Fsync) and applied on the filer's own HTTP
write path, but was never carried onto the chunk uploads S3 issues: the
AssignVolumeResponse had no fsync field, so the s3api client could not
learn the decision, and the chunked upload URL was hardcoded without it.
Every S3 write to a path with fsync configured went to the volume server
as a non-fsync write.

Carry the decision through the assign response:

- filer.proto: AssignVolumeResponse gains bool fsync, filled from the
  storage option the assign resolved.
- operation.AssignResult gains Fsync, so uploadChunk can append
  ?fsync=true to the volume server upload URL (single and replica
  fan-out paths).
- The S3 PUT/UploadPart assignFunc, the S3 copy path, the admin file
  browser upload, and the Iceberg worker assign functions all forward
  the response field.

Adds TestUploadReaderInChunksAppendsFsyncWhenAssigned.
2026-08-23 22:11:08 -07:00

270 lines
8.1 KiB
Go

package handlers
import (
"bytes"
"context"
"fmt"
"io"
"mime"
"net/http"
"path"
"strconv"
"strings"
"time"
"github.com/seaweedfs/seaweedfs/weed/admin/dash"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/operation"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/security"
)
// Admin upload chunk size, matching s3api so files split into the same fid-sized pieces.
const adminUploadChunkSize = 8 * 1024 * 1024
// File browser handlers backed by the filer gRPC service. They bypass the
// filer's HTTP listener so the UI keeps working when the filer is started
// with -disableHttp=true; chunk bytes still flow through the volume server
// HTTP endpoints (which run on their own ports).
// fetchFileContentGrpc reads file content via the filer gRPC service, looking
// the entry up and then streaming the chunks straight from the volume servers.
// When maxBytes > 0 the stream is truncated to that many bytes — used by the
// "is this text?" sniff so unknown-MIME files don't get fully downloaded.
func (h *FileBrowserHandlers) fetchFileContentGrpc(ctx context.Context, filePath string, maxBytes int) (string, error) {
cleanFilePath, err := h.validateAndCleanFilePath(filePath)
if err != nil {
return "", err
}
entry, err := h.lookupEntry(ctx, cleanFilePath)
if err != nil {
return "", err
}
if entry.IsDirectory {
return "", fmt.Errorf("path is a directory")
}
size := int64(filer.FileSize(entry))
streamSize := size
if maxBytes > 0 && streamSize > int64(maxBytes) {
streamSize = int64(maxBytes)
}
var buf bytes.Buffer
if err := h.streamEntryContent(ctx, entry, streamSize, &buf); err != nil {
return "", err
}
return buf.String(), nil
}
// downloadFileGrpc streams a file via gRPC + volume server HTTP. The
// response writer receives the canonical attachment headers and the raw
// bytes; this replaces the HTTP-to-filer proxy that used to run in
// DownloadFile.
func (h *FileBrowserHandlers) downloadFileGrpc(ctx context.Context, filePath string, w http.ResponseWriter, inline bool) error {
cleanFilePath, err := h.validateAndCleanFilePath(filePath)
if err != nil {
return err
}
entry, err := h.lookupEntry(ctx, cleanFilePath)
if err != nil {
return err
}
if entry.IsDirectory {
return fmt.Errorf("path is a directory")
}
size := int64(filer.FileSize(entry))
fileName := path.Base(cleanFilePath)
// Resolve mime like the viewer does so Content-Type and the inline check agree.
contentType := dash.ResolveEntryMime(entry)
// Only inline images and PDFs; serve the rest as attachments so a hostile upload
// (HTML, SVG) can't run as same-origin script. nosniff locks the declared type.
disposition := "attachment"
if inline && (strings.HasPrefix(contentType, "image/") || contentType == "application/pdf") {
disposition = "inline"
}
w.Header().Set("Content-Disposition", mime.FormatMediaType(disposition, map[string]string{"filename": fileName}))
w.Header().Set("Content-Type", contentType)
w.Header().Set("X-Content-Type-Options", "nosniff")
w.Header().Set("Content-Length", strconv.FormatInt(size, 10))
w.WriteHeader(http.StatusOK)
return h.streamEntryContent(ctx, entry, size, w)
}
// uploadFileGrpc streams the upload to volumes in 8 MiB chunks via the shared
// chunked-upload helper, then registers the entry over the filer gRPC service.
// Content always lands in volumes, never inlined on the entry.
func (h *FileBrowserHandlers) uploadFileGrpc(ctx context.Context, filePath string, fileName string, mimeType string, reader io.Reader) error {
cleanFilePath, err := h.validateAndCleanFilePath(filePath)
if err != nil {
return err
}
dir := path.Dir(cleanFilePath)
if dir == "." {
dir = "/"
}
entryName := path.Base(cleanFilePath)
if mimeType == "" {
mimeType = "application/octet-stream"
}
assignFunc := func(ctx context.Context, count int, expectedDataSize uint64) (*operation.VolumeAssignRequest, *operation.AssignResult, error) {
var assignResp *filer_pb.AssignVolumeResponse
err := h.adminServer.WithFilerClient(func(client filer_pb.SeaweedFilerClient) error {
resp, assignErr := client.AssignVolume(ctx, &filer_pb.AssignVolumeRequest{
Count: int32(count),
Path: cleanFilePath,
ExpectedDataSize: expectedDataSize,
})
if assignErr != nil {
return assignErr
}
if resp.Error != "" {
return fmt.Errorf("%s", resp.Error)
}
assignResp = resp
return nil
})
if err != nil {
return nil, nil, err
}
if assignResp.Location == nil || assignResp.FileId == "" {
return nil, nil, fmt.Errorf("assign volume returned empty location")
}
return nil, &operation.AssignResult{
Fid: assignResp.FileId,
Url: assignResp.Location.Url,
PublicUrl: assignResp.Location.PublicUrl,
Count: uint64(count),
Auth: security.EncodedJwt(assignResp.Auth),
Fsync: assignResp.Fsync,
}, nil
}
chunkResult, err := operation.UploadReaderInChunks(ctx, reader, &operation.ChunkedUploadOption{
ChunkSize: adminUploadChunkSize,
MimeType: mimeType,
AssignFunc: assignFunc,
})
if err != nil {
// Partial chunks come back even on error so we can clean them up rather
// than leaving orphaned data on volume servers.
if chunkResult != nil && len(chunkResult.FileChunks) > 0 {
h.deleteOrphanedChunks(chunkResult.FileChunks)
}
return fmt.Errorf("upload: %w", err)
}
now := time.Now()
entry := &filer_pb.Entry{
Name: entryName,
Attributes: &filer_pb.FuseAttributes{
FileSize: uint64(chunkResult.TotalSize),
Mtime: now.Unix(),
Crtime: now.Unix(),
FileMode: 0644,
Mime: mimeType,
},
}
entry.Chunks = chunkResult.FileChunks
err = h.adminServer.WithFilerClient(func(client filer_pb.SeaweedFilerClient) error {
_, createErr := client.CreateEntry(ctx, &filer_pb.CreateEntryRequest{
Directory: dir,
Entry: entry,
})
return createErr
})
if err != nil {
if len(chunkResult.FileChunks) > 0 {
h.deleteOrphanedChunks(chunkResult.FileChunks)
}
return fmt.Errorf("create entry: %w", err)
}
return nil
}
// deleteOrphanedChunks best-effort removes the chunk fids when an upload
// fails partway through. Errors are logged; we can't surface them past the
// caller's primary failure.
func (h *FileBrowserHandlers) deleteOrphanedChunks(chunks []*filer_pb.FileChunk) {
fileIds := make([]string, 0, len(chunks))
for _, c := range chunks {
if fid := c.GetFileIdString(); fid != "" {
fileIds = append(fileIds, fid)
}
}
if len(fileIds) == 0 {
return
}
master := h.adminServer.GetMasterClient()
results := operation.DeleteFileIds(master.GetMaster, false, h.adminServer.GetGrpcDialOption(), fileIds)
for _, r := range results {
if r.Error != "" {
glog.Warningf("admin file browser: orphan chunk %s cleanup: %s", r.FileId, r.Error)
}
}
}
func (h *FileBrowserHandlers) lookupEntry(ctx context.Context, cleanFilePath string) (*filer_pb.Entry, error) {
dir := path.Dir(cleanFilePath)
if dir == "." {
dir = "/"
}
name := path.Base(cleanFilePath)
var entry *filer_pb.Entry
err := h.adminServer.WithFilerClient(func(client filer_pb.SeaweedFilerClient) error {
resp, lookupErr := client.LookupDirectoryEntry(ctx, &filer_pb.LookupDirectoryEntryRequest{
Directory: dir,
Name: name,
})
if lookupErr != nil {
return lookupErr
}
if resp.Entry == nil {
return fmt.Errorf("not found")
}
entry = resp.Entry
return nil
})
return entry, err
}
func (h *FileBrowserHandlers) streamEntryContent(ctx context.Context, entry *filer_pb.Entry, size int64, w io.Writer) error {
if size == 0 {
// Inline content (small files stored directly on the entry) skip the
// chunk pipeline entirely.
if len(entry.Content) > 0 {
_, err := w.Write(entry.Content)
return err
}
return nil
}
if len(entry.Content) > 0 && len(entry.GetChunks()) == 0 {
_, err := w.Write(entry.Content)
return err
}
streamFn, err := filer.PrepareStreamContentWithThrottler(
ctx,
h.adminServer.GetMasterClient(),
dash.VolumeServerReadJwt,
entry.GetChunks(),
0,
size,
0,
)
if err != nil {
return fmt.Errorf("prepare stream: %w", err)
}
return streamFn(w)
}