Files
seaweedfs/weed/server/volume_server_handlers_read.go
T
acb44a68a9 fix(volume): skip unnecessary needle data reads on delete (#11711)
* fix(volume): do not read a needle's data to delete it

A delete read the needle's whole payload from disk only to discard it.
The delete handler and BatchDelete need the needle's cookie and its
chunk-manifest flag, which live in the header and trailer, but fetched
them with a full needle read, the same call a GET uses.

readNeedle now reads the header and trailer only when MustMetaOnly is
set, whatever the needle's size, and the delete paths ask for that. The
Rust volume server honours must_meta_only the same way for its delete
paths. A compressed needle or a chunk manifest still gets the full read,
as before, because their callers need the bytes.

HEAD is unchanged. It used to set MustMetaOnly, which nothing read; it
now asks for AttemptMetaOnly alone, so it keeps reading a needle up to
PagedReadLimit whole. On a spinning disk that one read is cheaper than a
header read followed by a trailer read.

Version 1 needles are always read in full. They have no data size field
and no metadata after the data, which ReadNeedleMeta relies on, so a
metadata-only read of one failed with "invalid needle meta size". That
already happened for version 1 needles above PagedReadLimit read with
AttemptMetaOnly.

A delete therefore no longer verifies the data CRC of the needle it
removes. The cookie check is unchanged.

* fix(volume): read transformed images fully on HEAD too

A HEAD carrying resize or crop params reports the transformed length,
which a metadata read cannot produce. Go and Rust both answered such
HEADs from the source header, returning the stored size instead of the
cropped or resized one.

Generated with [Devin](https://devin.ai)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>

---------

Co-authored-by: Chris Lu <chrislusf@users.noreply.github.com>
Co-authored-by: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-10-12 18:08:19 +08:00

483 lines
14 KiB
Go

package weed_server
import (
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"io"
"math/rand/v2"
"mime"
"net/http"
"net/url"
"path/filepath"
"strconv"
"strings"
"sync/atomic"
"time"
util_http "github.com/seaweedfs/seaweedfs/weed/util/http"
"github.com/seaweedfs/seaweedfs/weed/util/mem"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/images"
"github.com/seaweedfs/seaweedfs/weed/operation"
"github.com/seaweedfs/seaweedfs/weed/stats"
"github.com/seaweedfs/seaweedfs/weed/storage"
"github.com/seaweedfs/seaweedfs/weed/storage/erasure_coding"
"github.com/seaweedfs/seaweedfs/weed/storage/needle"
"github.com/seaweedfs/seaweedfs/weed/storage/types"
"github.com/seaweedfs/seaweedfs/weed/util"
)
const reqIsProxied = "proxied"
func NotFound(w http.ResponseWriter) {
stats.VolumeServerFileReadFailures.Inc()
stats.VolumeServerHandlerCounter.WithLabelValues(stats.ErrorGetNotFound).Inc()
w.WriteHeader(http.StatusNotFound)
}
func InternalError(w http.ResponseWriter) {
stats.VolumeServerFileReadFailures.Inc()
stats.VolumeServerHandlerCounter.WithLabelValues(stats.ErrorGetInternal).Inc()
w.WriteHeader(http.StatusInternalServerError)
}
func (vs *VolumeServer) proxyReqToTargetServer(w http.ResponseWriter, r *http.Request) {
vid, fid, _, _, _ := parseURLPath(r.URL.Path)
volumeId, err := needle.NewVolumeId(vid)
if err != nil {
glog.V(2).Infof("parsing vid %s: %v", r.URL.Path, err)
w.WriteHeader(http.StatusBadRequest)
return
}
lookupResult, err := operation.LookupVolumeId(vs.GetMaster, vs.grpcDialOption, volumeId.String())
if err != nil || len(lookupResult.Locations) <= 0 {
glog.V(0).Infoln("lookup error:", err, r.URL.Path)
stats.VolumeServerFileReadInvalidNeedles.Inc()
NotFound(w)
return
}
if len(lookupResult.Locations) >= 2 {
rand.Shuffle(len(lookupResult.Locations), func(i, j int) {
lookupResult.Locations[i], lookupResult.Locations[j] = lookupResult.Locations[j], lookupResult.Locations[i]
})
}
var targetUrl *url.URL
location := fmt.Sprintf("%s:%d", vs.store.Ip, vs.store.Port)
for _, loc := range lookupResult.Locations {
if !strings.Contains(loc.Url, location) {
rawURL, _ := util_http.NormalizeUrl(loc.Url)
targetUrl, _ = url.Parse(rawURL)
break
}
}
if targetUrl == nil {
stats.VolumeServerFileReadInvalidNeedles.Inc()
stats.VolumeServerHandlerCounter.WithLabelValues(stats.EmptyReadProxyLoc).Inc()
glog.Errorf("failed lookup target host is empty locations: %+v, %s", lookupResult.Locations, location)
NotFound(w)
return
}
if vs.ReadMode == "proxy" {
stats.VolumeServerHandlerCounter.WithLabelValues(stats.ReadProxyReq).Inc()
// proxy client request to target server
r.URL.Host = targetUrl.Host
r.URL.Scheme = targetUrl.Scheme
query := r.URL.Query()
query.Set(reqIsProxied, "true")
r.URL.RawQuery = query.Encode()
request, err := http.NewRequest(http.MethodGet, r.URL.String(), nil)
if err != nil {
glog.V(0).Infof("failed to instance http request of url %s: %v", r.URL.String(), err)
InternalError(w)
return
}
for k, vv := range r.Header {
for _, v := range vv {
request.Header.Add(k, v)
}
}
response, err := util_http.GetGlobalHttpClient().Do(request)
if err != nil {
stats.VolumeServerHandlerCounter.WithLabelValues(stats.FailedReadProxyReq).Inc()
glog.V(0).Infof("request remote url %s: %v", r.URL.String(), err)
InternalError(w)
return
}
defer util_http.CloseResponse(response)
// proxy target response to client
for k, vv := range response.Header {
if k == "Server" {
continue
}
for _, v := range vv {
w.Header().Add(k, v)
}
}
w.WriteHeader(response.StatusCode)
buf := mem.Allocate(128 * 1024)
defer mem.Free(buf)
io.CopyBuffer(w, response.Body, buf)
return
} else {
// redirect
stats.VolumeServerHandlerCounter.WithLabelValues(stats.ReadRedirectReq).Inc()
targetUrl.Path = fmt.Sprintf("%s/%s,%s", targetUrl.Path, vid, fid)
arg := url.Values{}
if c := r.FormValue("collection"); c != "" {
arg.Set("collection", c)
}
arg.Set(reqIsProxied, "true")
targetUrl.RawQuery = arg.Encode()
http.Redirect(w, r, targetUrl.String(), http.StatusMovedPermanently)
return
}
}
func (vs *VolumeServer) GetOrHeadHandler(w http.ResponseWriter, r *http.Request) {
n := new(needle.Needle)
vid, fid, filename, ext, _ := parseURLPath(r.URL.Path)
if !vs.maybeCheckJwtAuthorization(r, vid, fid, false) {
writeJsonError(w, r, http.StatusUnauthorized, errors.New("wrong jwt"))
return
}
volumeId, err := needle.NewVolumeId(vid)
if err != nil {
glog.V(2).Infof("parsing vid %s: %v", r.URL.Path, err)
w.WriteHeader(http.StatusBadRequest)
return
}
err = n.ParsePath(fid)
if err != nil {
glog.V(2).Infof("parsing fid %s: %v", r.URL.Path, err)
w.WriteHeader(http.StatusBadRequest)
return
}
// glog.V(4).Infoln("volume", volumeId, "reading", n)
hasVolume := vs.store.HasVolume(volumeId)
_, hasEcVolume := vs.store.FindEcVolume(volumeId)
if !hasVolume && !hasEcVolume {
if vs.ReadMode == "local" {
stats.VolumeServerFileReadInvalidNeedles.Inc()
glog.V(0).Infoln("volume is not local:", err, r.URL.Path)
NotFound(w)
return
}
vs.proxyReqToTargetServer(w, r)
return
}
cookie := n.Cookie
readOption := &storage.ReadOption{
ReadDeleted: r.FormValue("readDeleted") == "true",
HasSlowRead: vs.hasSlowRead,
ReadBufferSize: vs.readBufferSizeMB * 1024 * 1024,
}
var count int
var memoryCost types.Size
readOption.AttemptMetaOnly, readOption.MustMetaOnly = shouldAttemptStreamWrite(hasVolume, ext, r)
onReadSizeFn := func(size types.Size) {
memoryCost = size
atomic.AddInt64(&vs.inFlightDownloadDataSize, int64(memoryCost))
}
if hasVolume {
count, err = vs.store.ReadVolumeNeedle(volumeId, n, readOption, onReadSizeFn)
} else if hasEcVolume {
count, err = vs.store.ReadEcShardNeedle(volumeId, n, onReadSizeFn)
}
defer func() {
atomic.AddInt64(&vs.inFlightDownloadDataSize, -int64(memoryCost))
if vs.concurrentDownloadLimit != 0 {
vs.inFlightDownloadDataLimitCond.Broadcast()
}
}()
if err != nil && err != storage.ErrorDeleted && hasVolume {
glog.V(4).Infof("read needle: %v", err)
// start to fix it from other replicas, if not deleted and hasVolume and is not a replicated request
}
// glog.V(4).Infoln("read bytes", count, "error", err)
if err != nil || count < 0 {
glog.V(3).Infof("read %s isNormalVolume %v error: %v", r.URL.Path, hasVolume, err)
invalid_needle := err == storage.ErrorNotFound || errors.Is(err, erasure_coding.NotFoundError)
if invalid_needle {
stats.VolumeServerFileReadInvalidNeedles.Inc()
}
if invalid_needle || err == storage.ErrorDeleted {
NotFound(w)
} else {
InternalError(w)
}
return
}
if n.Cookie != cookie {
glog.V(0).Infof("request %s with cookie:%x expected:%x from %s agent %s", r.URL.Path, cookie, n.Cookie, r.RemoteAddr, r.UserAgent())
NotFound(w)
return
}
if n.LastModified != 0 {
w.Header().Set("Last-Modified", time.Unix(int64(n.LastModified), 0).UTC().Format(http.TimeFormat))
if r.Header.Get("If-Modified-Since") != "" {
if t, parseError := time.Parse(http.TimeFormat, r.Header.Get("If-Modified-Since")); parseError == nil {
if t.Unix() >= int64(n.LastModified) {
w.WriteHeader(http.StatusNotModified)
return
}
}
}
}
if inm := r.Header.Get("If-None-Match"); inm == "\""+n.Etag()+"\"" {
w.WriteHeader(http.StatusNotModified)
return
}
SetEtag(w, n.Etag())
if n.HasPairs() {
pairMap := make(map[string]string)
err = json.Unmarshal(n.Pairs, &pairMap)
if err != nil {
glog.V(0).Infoln("Unmarshal pairs error:", err)
}
for k, v := range pairMap {
w.Header().Set(k, v)
}
}
if vs.tryHandleChunkedFile(n, filename, ext, w, r) {
return
}
if n.NameSize > 0 && filename == "" {
filename = string(n.Name)
if ext == "" {
ext = filepath.Ext(filename)
}
}
mtype := ""
if n.MimeSize > 0 {
mt := string(n.Mime)
if !strings.HasPrefix(mt, "application/octet-stream") {
mtype = mt
}
}
if n.IsCompressed() {
_, _, _, shouldResize := shouldResizeImages(ext, r)
_, _, _, _, shouldCrop := shouldCropImages(ext, r)
if shouldResize || shouldCrop {
if n.Data, err = util.DecompressData(n.Data); err != nil {
glog.V(0).Infoln("ungzip error:", err, r.URL.Path)
}
// } else if strings.Contains(r.Header.Get("Accept-Encoding"), "zstd") && util.IsZstdContent(n.Data) {
// w.Header().Set("Content-Encoding", "zstd")
} else if strings.Contains(r.Header.Get("Accept-Encoding"), "gzip") && util.IsGzippedContent(n.Data) {
w.Header().Set("Content-Encoding", "gzip")
} else {
if n.Data, err = util.DecompressData(n.Data); err != nil {
glog.V(0).Infoln("uncompress error:", err, r.URL.Path)
}
}
}
if !readOption.IsMetaOnly {
rs := conditionallyCropImages(bytes.NewReader(n.Data), ext, r)
rs = conditionallyResizeImages(rs, ext, r)
if e := writeResponseContent(filename, mtype, rs, w, r); e != nil {
glog.V(2).Infoln("response write error:", e)
}
} else {
vs.streamWriteResponseContent(filename, mtype, volumeId, n, w, r, readOption)
}
}
func shouldAttemptStreamWrite(hasLocalVolume bool, ext string, r *http.Request) (shouldAttempt bool, mustMetaOnly bool) {
if !hasLocalVolume {
return false, false
}
if len(ext) > 0 {
ext = strings.ToLower(ext)
}
_, _, _, shouldResize := shouldResizeImages(ext, r)
_, _, _, _, shouldCrop := shouldCropImages(ext, r)
if shouldResize || shouldCrop {
// A transformed image's length is not the needle's; the payload is
// needed to answer — also true for HEAD, which reports it.
return false, false
}
if r.Method == http.MethodHead {
// Attempt only, so a needle up to PagedReadLimit is still read whole:
// on a spinning disk one read is cheaper than a header read followed
// by a trailer read.
return true, false
}
return true, false
}
func (vs *VolumeServer) tryHandleChunkedFile(n *needle.Needle, fileName string, ext string, w http.ResponseWriter, r *http.Request) (processed bool) {
if !n.IsChunkedManifest() || r.URL.Query().Get("cm") == "false" {
return false
}
chunkManifest, e := operation.LoadChunkManifest(n.Data, n.IsCompressed())
if e != nil {
glog.V(0).Infof("load chunked manifest (%s) error: %v", r.URL.Path, e)
return false
}
if fileName == "" && chunkManifest.Name != "" {
fileName = chunkManifest.Name
}
if ext == "" {
ext = filepath.Ext(fileName)
}
mType := ""
if chunkManifest.Mime != "" {
mt := chunkManifest.Mime
if !strings.HasPrefix(mt, "application/octet-stream") {
mType = mt
}
}
w.Header().Set("X-File-Store", "chunked")
chunkedFileReader := operation.NewChunkedFileReader(chunkManifest.Chunks, vs.GetMaster(context.Background()), vs.grpcDialOption)
defer chunkedFileReader.Close()
rs := conditionallyCropImages(chunkedFileReader, ext, r)
rs = conditionallyResizeImages(rs, ext, r)
if e := writeResponseContent(fileName, mType, rs, w, r); e != nil {
glog.V(2).Infoln("response write error:", e)
}
return true
}
func conditionallyResizeImages(originalDataReaderSeeker io.ReadSeeker, ext string, r *http.Request) io.ReadSeeker {
rs := originalDataReaderSeeker
if len(ext) > 0 {
ext = strings.ToLower(ext)
}
width, height, mode, shouldResize := shouldResizeImages(ext, r)
if shouldResize {
rs, _, _ = images.Resized(ext, originalDataReaderSeeker, width, height, mode)
}
return rs
}
func shouldResizeImages(ext string, r *http.Request) (width, height int, mode string, shouldResize bool) {
if ext == ".png" || ext == ".jpg" || ext == ".jpeg" || ext == ".gif" || ext == ".webp" {
if r.FormValue("width") != "" {
width, _ = strconv.Atoi(r.FormValue("width"))
}
if r.FormValue("height") != "" {
height, _ = strconv.Atoi(r.FormValue("height"))
}
}
mode = r.FormValue("mode")
shouldResize = width > 0 || height > 0
return
}
func conditionallyCropImages(originalDataReaderSeeker io.ReadSeeker, ext string, r *http.Request) io.ReadSeeker {
rs := originalDataReaderSeeker
if len(ext) > 0 {
ext = strings.ToLower(ext)
}
x1, y1, x2, y2, shouldCrop := shouldCropImages(ext, r)
if shouldCrop {
var err error
rs, err = images.Cropped(ext, rs, x1, y1, x2, y2)
if err != nil {
glog.Errorf("Cropping images error: %s", err)
}
}
return rs
}
func shouldCropImages(ext string, r *http.Request) (x1, y1, x2, y2 int, shouldCrop bool) {
if ext == ".png" || ext == ".jpg" || ext == ".jpeg" || ext == ".gif" {
if r.FormValue("crop_x1") != "" {
x1, _ = strconv.Atoi(r.FormValue("crop_x1"))
}
if r.FormValue("crop_y1") != "" {
y1, _ = strconv.Atoi(r.FormValue("crop_y1"))
}
if r.FormValue("crop_x2") != "" {
x2, _ = strconv.Atoi(r.FormValue("crop_x2"))
}
if r.FormValue("crop_y2") != "" {
y2, _ = strconv.Atoi(r.FormValue("crop_y2"))
}
}
shouldCrop = x1 >= 0 && y1 >= 0 && x2 > x1 && y2 > y1
return
}
func writeResponseContent(filename, mimeType string, rs io.ReadSeeker, w http.ResponseWriter, r *http.Request) error {
totalSize, e := rs.Seek(0, 2)
if mimeType == "" {
if ext := filepath.Ext(filename); ext != "" {
mimeType = mime.TypeByExtension(ext)
}
}
if mimeType != "" {
w.Header().Set("Content-Type", mimeType)
}
w.Header().Set("Accept-Ranges", "bytes")
AdjustPassthroughHeaders(w, r, filename)
if r.Method == http.MethodHead {
w.Header().Set("Content-Length", strconv.FormatInt(totalSize, 10))
return nil
}
return ProcessRangeRequest(r, w, totalSize, mimeType, func(offset int64, size int64) (filer.DoStreamContent, error) {
return func(writer io.Writer) error {
if _, e = rs.Seek(offset, 0); e != nil {
return e
}
_, e = io.CopyN(writer, rs, size)
return e
}, nil
})
}
func (vs *VolumeServer) streamWriteResponseContent(filename string, mimeType string, volumeId needle.VolumeId, n *needle.Needle, w http.ResponseWriter, r *http.Request, readOption *storage.ReadOption) {
totalSize := int64(n.DataSize)
if mimeType == "" {
if ext := filepath.Ext(filename); ext != "" {
mimeType = mime.TypeByExtension(ext)
}
}
if mimeType != "" {
w.Header().Set("Content-Type", mimeType)
}
w.Header().Set("Accept-Ranges", "bytes")
AdjustPassthroughHeaders(w, r, filename)
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) {
return func(writer io.Writer) error {
return vs.store.ReadVolumeNeedleDataInto(volumeId, n, readOption, writer, offset, size)
}, nil
})
}