Files
seaweedfs/weed/server/filer_server_handlers_read.go
T
Chris Lu 7f770b1553 fix(filer): return 503 + Retry-After when remote object not cached yet (#9236)
* extend cache-not-ready handling to filer HTTP path

Mirror the s3api change for the native filer HTTP handlers. When the
filer GET hits a remote-only object whose cache fill hasn't completed,
return 503 Service Unavailable with Retry-After: 5 instead of 500
Internal Error, and treat client disconnects as silent cancellations
rather than logging them as errors.

Adds an ErrCacheNotReady sentinel and a small helper used at the
prepareWriteFn-error sites in ProcessRangeRequest, so the same
classification (cancel / not-ready / other) applies to plain GETs,
single-range, and multi-range requests.

* clear Content-Range on prepareWriteFn error

The single-range path sets Content-Range before calling prepareWriteFn.
If prepareWriteFn fails, http.Error is about to write a fresh body for
503 or 500, but the stale Content-Range header would still go out and
no longer match. Drop it alongside Content-Length in the shared helper
so all current and future callers are covered.

* strip success-path headers and forward NotFound on prepareWriteFn error

When ProcessRangeRequest writes an error response, the previously-set
success headers (Content-Disposition, ETag, Last-Modified, in addition
to Content-Length/Content-Range) shouldn't ride along on the new body.
With ?dl=1 a stale Content-Disposition would even cause browsers to
save the error message under the object's filename. Strip them all in
the shared helper.

Also forward filer_pb.ErrNotFound through the cache-failure branch so a
mid-cache entry deletion surfaces as 404, not as a 503 retry-loop.
Permanent upstream cloud errors (403/404 from the cloud SDK) still come
back as opaque wrapped strings via FetchAndWriteNeedle and remain
mapped to 503; distinguishing those would need a wider refactor.
2026-04-27 01:58:33 -07:00

259 lines
8.4 KiB
Go

package weed_server
import (
"context"
"errors"
"fmt"
"io"
"math"
"mime"
"net/http"
"path/filepath"
"strconv"
"strings"
"time"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3_constants"
"github.com/seaweedfs/seaweedfs/weed/security"
"github.com/seaweedfs/seaweedfs/weed/stats"
"github.com/seaweedfs/seaweedfs/weed/util"
)
// Validates the preconditions. Returns true if GET/HEAD operation should not proceed.
// Preconditions supported are:
//
// If-Modified-Since
// If-Unmodified-Since
// If-Match
// If-None-Match
func checkPreconditions(w http.ResponseWriter, r *http.Request, entry *filer.Entry) bool {
etag := filer.ETagEntry(entry)
/// When more than one conditional request header field is present in a
/// request, the order in which the fields are evaluated becomes
/// important. In practice, the fields defined in this document are
/// consistently implemented in a single, logical order, since "lost
/// update" preconditions have more strict requirements than cache
/// validation, a validated cache is more efficient than a partial
/// response, and entity tags are presumed to be more accurate than date
/// validators. https://tools.ietf.org/html/rfc7232#section-5
if entry.Attr.Mtime.IsZero() {
return false
}
w.Header().Set("Last-Modified", entry.Attr.Mtime.UTC().Format(http.TimeFormat))
ifMatchETagHeader := r.Header.Get("If-Match")
ifUnmodifiedSinceHeader := r.Header.Get("If-Unmodified-Since")
if ifMatchETagHeader != "" {
if util.CanonicalizeETag(etag) != util.CanonicalizeETag(ifMatchETagHeader) {
w.WriteHeader(http.StatusPreconditionFailed)
return true
}
} else if ifUnmodifiedSinceHeader != "" {
if t, parseError := time.Parse(http.TimeFormat, ifUnmodifiedSinceHeader); parseError == nil {
if t.Before(entry.Attr.Mtime) {
w.WriteHeader(http.StatusPreconditionFailed)
return true
}
}
}
ifNoneMatchETagHeader := r.Header.Get("If-None-Match")
ifModifiedSinceHeader := r.Header.Get("If-Modified-Since")
if ifNoneMatchETagHeader != "" {
if util.CanonicalizeETag(etag) == util.CanonicalizeETag(ifNoneMatchETagHeader) {
SetEtag(w, etag)
w.WriteHeader(http.StatusNotModified)
return true
}
} else if ifModifiedSinceHeader != "" {
if t, parseError := time.Parse(http.TimeFormat, ifModifiedSinceHeader); parseError == nil {
if !t.Before(entry.Attr.Mtime) {
SetEtag(w, etag)
w.WriteHeader(http.StatusNotModified)
return true
}
}
}
return false
}
func (fs *FilerServer) GetOrHeadHandler(w http.ResponseWriter, r *http.Request) {
ctx := r.Context()
path := r.URL.Path
isForDirectory := strings.HasSuffix(path, "/")
if isForDirectory && len(path) > 1 {
path = path[:len(path)-1]
}
entry, err := fs.filer.FindEntry(ctx, util.FullPath(path))
if err != nil {
if path == "/" {
fs.listDirectoryHandler(w, r)
return
}
if err == filer_pb.ErrNotFound {
glog.V(2).InfofCtx(ctx, "Not found %s: %v", path, err)
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorReadNotFound).Inc()
w.WriteHeader(http.StatusNotFound)
} else {
glog.ErrorfCtx(ctx, "Internal %s: %v", path, err)
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorReadInternal).Inc()
w.WriteHeader(http.StatusInternalServerError)
}
return
}
query := r.URL.Query()
if entry.IsDirectory() {
if fs.option.DisableDirListing {
w.WriteHeader(http.StatusForbidden)
return
}
if query.Get("metadata") == "true" {
writeJsonQuiet(w, r, http.StatusOK, entry)
return
}
// listDirectoryHandler checks ExposeDirectoryData internally
fs.listDirectoryHandler(w, r)
return
}
if query.Get("metadata") == "true" {
if query.Get("resolveManifest") == "true" {
if entry.Chunks, _, err = filer.ResolveChunkManifest(
ctx,
fs.filer.MasterClient.GetLookupFileIdFunction(),
entry.GetChunks(), 0, math.MaxInt64); err != nil {
err = fmt.Errorf("failed to resolve chunk manifest, err: %s", err.Error())
writeJsonError(w, r, http.StatusInternalServerError, err)
return
}
}
writeJsonQuiet(w, r, http.StatusOK, entry)
return
}
if checkPreconditions(w, r, entry) {
return
}
// Generate ETag for response
etag := filer.ETagEntry(entry)
w.Header().Set("Accept-Ranges", "bytes")
// mime type
mimeType := entry.Attr.Mime
if mimeType == "" {
if ext := filepath.Ext(entry.Name()); ext != "" {
mimeType = mime.TypeByExtension(ext)
}
}
if mimeType != "" {
w.Header().Set("Content-Type", mimeType)
} else {
w.Header().Set("Content-Type", "application/octet-stream")
}
// print out the header from extended properties
// Filter out xattr-* (filesystem extended attributes) and internal SeaweedFS headers
for k, v := range entry.Extended {
if !strings.HasPrefix(k, "xattr-") && !s3_constants.IsSeaweedFSInternalHeader(k) {
w.Header().Set(k, string(v))
}
}
//Seaweed custom header are not visible to Vue or javascript
seaweedHeaders := []string{}
for header := range w.Header() {
if strings.HasPrefix(header, "Seaweed-") {
seaweedHeaders = append(seaweedHeaders, header)
}
}
seaweedHeaders = append(seaweedHeaders, "Content-Disposition")
w.Header().Set("Access-Control-Expose-Headers", strings.Join(seaweedHeaders, ","))
SetEtag(w, etag)
filename := entry.Name()
AdjustPassthroughHeaders(w, r, filename)
// For range processing, use the original content size, not the encrypted size
// entry.Size() returns max(chunk_sizes, file_size) where chunk_sizes include encryption overhead
// For SSE objects, we need the original unencrypted size for proper range validation
totalSize := int64(entry.FileSize)
if r.Method == http.MethodHead {
w.Header().Set("Content-Length", strconv.FormatInt(totalSize, 10))
return
}
ProcessRangeRequest(r, w, totalSize, mimeType, func(offset int64, size int64) (filer.DoStreamContent, error) {
if offset+size <= int64(len(entry.Content)) {
return func(writer io.Writer) error {
_, err := writer.Write(entry.Content[offset : offset+size])
if err != nil {
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorWriteEntry).Inc()
glog.ErrorfCtx(ctx, "failed to write entry content: %v", err)
}
return err
}, nil
}
chunks := entry.GetChunks()
if entry.IsInRemoteOnly() {
dir, name := entry.FullPath.DirAndName()
if resp, err := fs.CacheRemoteObjectToLocalCluster(ctx, &filer_pb.CacheRemoteObjectToLocalClusterRequest{
Directory: dir,
Name: name,
}); err != nil {
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorReadCache).Inc()
// Client disconnected: surface ctx error so caller stays silent.
if ctxErr := ctx.Err(); ctxErr != nil {
return nil, ctxErr
}
// Entry vanished mid-cache: forward NotFound so caller maps to 404,
// not the 503 retry-loop.
if errors.Is(err, filer_pb.ErrNotFound) {
return nil, err
}
// Cache still filling: tag with sentinel so caller maps to 503 + Retry-After.
glog.WarningfCtx(ctx, "CacheRemoteObjectToLocalCluster %s: %v", entry.FullPath, err)
return nil, fmt.Errorf("cache %s: %w", entry.FullPath, ErrCacheNotReady)
} else {
chunks = resp.Entry.GetChunks()
}
}
// Use a detached context for streaming so client disconnects/cancellations don't abort volume server operations,
// while preserving request-scoped values like tracing IDs.
// Matches S3 API behavior. Request context (ctx) is used for metadata operations above.
streamCtx, streamCancel := context.WithCancel(context.WithoutCancel(ctx))
streamFn, err := filer.PrepareStreamContentWithPrefetch(streamCtx, fs.filer.MasterClient, fs.maybeGetVolumeReadJwtAuthorizationToken, chunks, offset, size, fs.option.DownloadMaxBytesPs, 4)
if err != nil {
streamCancel()
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorReadStream).Inc()
glog.ErrorfCtx(ctx, "failed to prepare stream content %s: %v", r.URL, err)
return nil, err
}
return func(writer io.Writer) error {
defer streamCancel()
err := streamFn(writer)
if err != nil {
stats.FilerHandlerCounter.WithLabelValues(stats.ErrorReadStream).Inc()
glog.ErrorfCtx(ctx, "failed to stream content %s: %v", r.URL, err)
}
return err
}, nil
})
}
func (fs *FilerServer) maybeGetVolumeReadJwtAuthorizationToken(fileId string) string {
return string(security.GenJwtForVolumeServer(fs.volumeGuard.ReadSigningKey, fs.volumeGuard.ReadExpiresAfterSec, fileId))
}