Browse Source

better control for reader caching

original_weed_mount
chrislu 3 years ago
parent
commit
28b395bef4
  1. 30
      weed/filer/filechunk_manifest.go
  2. 130
      weed/filer/reader_at.go
  3. 17
      weed/filer/reader_at_test.go
  4. 187
      weed/filer/reader_cache.go
  5. 18
      weed/filer/stream.go
  6. 5
      weed/server/filer_server_handlers_read.go
  7. 103
      weed/util/chunk_cache/chunk_cache.go

30
weed/filer/filechunk_manifest.go

@ -3,6 +3,7 @@ package filer
import (
"bytes"
"fmt"
"github.com/chrislusf/seaweedfs/weed/util/mem"
"github.com/chrislusf/seaweedfs/weed/wdclient"
"io"
"math"
@ -77,7 +78,9 @@ func ResolveOneChunkManifest(lookupFileIdFn wdclient.LookupFileIdFunctionType, c
}
// IsChunkManifest
data, err := fetchChunk(lookupFileIdFn, chunk.GetFileIdString(), chunk.CipherKey, chunk.IsCompressed)
data := mem.Allocate(int(chunk.Size))
defer mem.Free(data)
_, err := fetchChunk(data, lookupFileIdFn, chunk.GetFileIdString(), chunk.CipherKey, chunk.IsCompressed)
if err != nil {
return nil, fmt.Errorf("fail to read manifest %s: %v", chunk.GetFileIdString(), err)
}
@ -92,38 +95,37 @@ func ResolveOneChunkManifest(lookupFileIdFn wdclient.LookupFileIdFunctionType, c
}
// TODO fetch from cache for weed mount?
func fetchChunk(lookupFileIdFn wdclient.LookupFileIdFunctionType, fileId string, cipherKey []byte, isGzipped bool) ([]byte, error) {
func fetchChunk(data []byte, lookupFileIdFn wdclient.LookupFileIdFunctionType, fileId string, cipherKey []byte, isGzipped bool) (int, error) {
urlStrings, err := lookupFileIdFn(fileId)
if err != nil {
glog.Errorf("operation LookupFileId %s failed, err: %v", fileId, err)
return nil, err
return 0, err
}
return retriedFetchChunkData(urlStrings, cipherKey, isGzipped, true, 0, 0)
return retriedFetchChunkData(data, urlStrings, cipherKey, isGzipped, true, 0)
}
func fetchChunkRange(lookupFileIdFn wdclient.LookupFileIdFunctionType, fileId string, cipherKey []byte, isGzipped bool, offset int64, size int) ([]byte, error) {
func fetchChunkRange(data []byte, lookupFileIdFn wdclient.LookupFileIdFunctionType, fileId string, cipherKey []byte, isGzipped bool, offset int64) (int, error) {
urlStrings, err := lookupFileIdFn(fileId)
if err != nil {
glog.Errorf("operation LookupFileId %s failed, err: %v", fileId, err)
return nil, err
return 0, err
}
return retriedFetchChunkData(urlStrings, cipherKey, isGzipped, false, offset, size)
return retriedFetchChunkData(data, urlStrings, cipherKey, isGzipped, false, offset)
}
func retriedFetchChunkData(urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64, size int) ([]byte, error) {
func retriedFetchChunkData(buffer []byte, urlStrings []string, cipherKey []byte, isGzipped bool, isFullChunk bool, offset int64) (n int, err error) {
var err error
var shouldRetry bool
receivedData := make([]byte, 0, size)
for waitTime := time.Second; waitTime < util.RetryWaitTime; waitTime += waitTime / 2 {
for _, urlString := range urlStrings {
receivedData = receivedData[:0]
n = 0
if strings.Contains(urlString, "%") {
urlString = url.PathEscape(urlString)
}
shouldRetry, err = util.ReadUrlAsStream(urlString+"?readDeleted=true", cipherKey, isGzipped, isFullChunk, offset, size, func(data []byte) {
receivedData = append(receivedData, data...)
shouldRetry, err = util.ReadUrlAsStream(urlString+"?readDeleted=true", cipherKey, isGzipped, isFullChunk, offset, len(buffer), func(data []byte) {
x := copy(buffer[n:], data)
n += x
})
if !shouldRetry {
break
@ -142,7 +144,7 @@ func retriedFetchChunkData(urlStrings []string, cipherKey []byte, isGzipped bool
}
}
return receivedData, err
return n, err
}

130
weed/filer/reader_at.go

@ -12,21 +12,15 @@ import (
"github.com/chrislusf/seaweedfs/weed/util"
"github.com/chrislusf/seaweedfs/weed/util/chunk_cache"
"github.com/chrislusf/seaweedfs/weed/wdclient"
"github.com/golang/groupcache/singleflight"
)
type ChunkReadAt struct {
masterClient *wdclient.MasterClient
chunkViews []*ChunkView
lookupFileId wdclient.LookupFileIdFunctionType
readerLock sync.Mutex
fileSize int64
fetchGroup singleflight.Group
chunkCache chunk_cache.ChunkCache
lastChunkFileId string
lastChunkData []byte
readerPattern *ReaderPattern
masterClient *wdclient.MasterClient
chunkViews []*ChunkView
readerLock sync.Mutex
fileSize int64
readerCache *ReaderCache
readerPattern *ReaderPattern
}
var _ = io.ReaderAt(&ChunkReadAt{})
@ -90,16 +84,14 @@ func NewChunkReaderAtFromClient(lookupFn wdclient.LookupFileIdFunctionType, chun
return &ChunkReadAt{
chunkViews: chunkViews,
lookupFileId: lookupFn,
chunkCache: chunkCache,
fileSize: fileSize,
readerCache: newReaderCache(5, chunkCache, lookupFn),
readerPattern: NewReaderPattern(),
}
}
func (c *ChunkReadAt) Close() error {
c.lastChunkData = nil
c.lastChunkFileId = ""
c.readerCache.destroy()
return nil
}
@ -142,16 +134,13 @@ func (c *ChunkReadAt) doReadAt(p []byte, offset int64) (n int, err error) {
continue
}
// glog.V(4).Infof("read [%d,%d), %d/%d chunk %s [%d,%d)", chunkStart, chunkStop, i, len(c.chunkViews), chunk.FileId, chunk.LogicOffset-chunk.Offset, chunk.LogicOffset-chunk.Offset+int64(chunk.Size))
var buffer []byte
bufferOffset := chunkStart - chunk.LogicOffset + chunk.Offset
bufferLength := chunkStop - chunkStart
buffer, err = c.readChunkSlice(chunk, nextChunk, uint64(bufferOffset), uint64(bufferLength))
copied, err := c.readChunkSliceAt(p[startOffset-offset:chunkStop-chunkStart+startOffset-offset], chunk, nextChunk, uint64(bufferOffset))
if err != nil {
glog.Errorf("fetching chunk %+v: %v\n", chunk, err)
return
return copied, err
}
copied := copy(p[startOffset-offset:chunkStop-chunkStart+startOffset-offset], buffer)
n += copied
startOffset, remaining = startOffset+int64(copied), remaining-int64(copied)
}
@ -173,104 +162,15 @@ func (c *ChunkReadAt) doReadAt(p []byte, offset int64) (n int, err error) {
}
func (c *ChunkReadAt) readChunkSlice(chunkView *ChunkView, nextChunkViews *ChunkView, offset, length uint64) ([]byte, error) {
func (c *ChunkReadAt) readChunkSliceAt(buffer []byte, chunkView *ChunkView, nextChunkViews *ChunkView, offset uint64) (n int, err error) {
var chunkSlice []byte
if chunkView.LogicOffset == 0 {
chunkSlice = c.chunkCache.GetChunkSlice(chunkView.FileId, offset, length)
}
if len(chunkSlice) > 0 {
return chunkSlice, nil
}
if c.lookupFileId == nil {
return nil, nil
}
if c.readerPattern.IsRandomMode() {
return c.doFetchRangeChunkData(chunkView, offset, length)
}
chunkData, err := c.readFromWholeChunkData(chunkView, nextChunkViews)
if err != nil {
return nil, err
return c.readerCache.ReadChunkAt(buffer, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset), int(chunkView.ChunkSize), chunkView.LogicOffset == 0)
}
wanted := min(int64(length), int64(len(chunkData))-int64(offset))
return chunkData[offset : int64(offset)+wanted], nil
}
func (c *ChunkReadAt) readFromWholeChunkData(chunkView *ChunkView, nextChunkViews ...*ChunkView) (chunkData []byte, err error) {
if c.lastChunkFileId == chunkView.FileId {
return c.lastChunkData, nil
n, err = c.readerCache.ReadChunkAt(buffer, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset), int(chunkView.ChunkSize), chunkView.LogicOffset == 0)
if nextChunkViews != nil {
c.readerCache.MaybeCache(nextChunkViews.FileId, nextChunkViews.CipherKey, nextChunkViews.IsGzipped, int(nextChunkViews.ChunkSize))
}
v, doErr := c.readOneWholeChunk(chunkView)
if doErr != nil {
return nil, doErr
}
chunkData = v.([]byte)
c.lastChunkData = chunkData
c.lastChunkFileId = chunkView.FileId
for _, nextChunkView := range nextChunkViews {
if c.chunkCache != nil && nextChunkView != nil {
go c.readOneWholeChunk(nextChunkView)
}
}
return
}
func (c *ChunkReadAt) readOneWholeChunk(chunkView *ChunkView) (interface{}, error) {
var err error
return c.fetchGroup.Do(chunkView.FileId, func() (interface{}, error) {
glog.V(4).Infof("readFromWholeChunkData %s offset %d [%d,%d) size at least %d", chunkView.FileId, chunkView.Offset, chunkView.LogicOffset, chunkView.LogicOffset+int64(chunkView.Size), chunkView.ChunkSize)
var data []byte
if chunkView.LogicOffset == 0 {
data = c.chunkCache.GetChunk(chunkView.FileId, chunkView.ChunkSize)
}
if data != nil {
glog.V(4).Infof("cache hit %s [%d,%d)", chunkView.FileId, chunkView.LogicOffset-chunkView.Offset, chunkView.LogicOffset-chunkView.Offset+int64(len(data)))
} else {
var err error
data, err = c.doFetchFullChunkData(chunkView)
if err != nil {
return data, err
}
if chunkView.LogicOffset == 0 {
// only cache the first chunk
c.chunkCache.SetChunk(chunkView.FileId, data)
}
}
return data, err
})
}
func (c *ChunkReadAt) doFetchFullChunkData(chunkView *ChunkView) ([]byte, error) {
glog.V(4).Infof("+ doFetchFullChunkData %s", chunkView.FileId)
data, err := fetchChunk(c.lookupFileId, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped)
glog.V(4).Infof("- doFetchFullChunkData %s", chunkView.FileId)
return data, err
}
func (c *ChunkReadAt) doFetchRangeChunkData(chunkView *ChunkView, offset, length uint64) ([]byte, error) {
glog.V(4).Infof("+ doFetchFullChunkData %s", chunkView.FileId)
data, err := fetchChunkRange(c.lookupFileId, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset), int(length))
glog.V(4).Infof("- doFetchFullChunkData %s", chunkView.FileId)
return data, err
}

17
weed/filer/reader_at_test.go

@ -21,8 +21,12 @@ func (m *mockChunkCache) GetChunk(fileId string, minSize uint64) (data []byte) {
return data
}
func (m *mockChunkCache) GetChunkSlice(fileId string, offset, length uint64) []byte {
return m.GetChunk(fileId, length)
func (m *mockChunkCache) ReadChunkAt(data []byte, fileId string, offset uint64) (n int, err error) {
x, _ := strconv.Atoi(fileId)
for i := 0; i < len(data); i++ {
data[i] = byte(x)
}
return len(data), nil
}
func (m *mockChunkCache) SetChunk(fileId string, data []byte) {
@ -65,10 +69,9 @@ func TestReaderAt(t *testing.T) {
readerAt := &ChunkReadAt{
chunkViews: ViewFromVisibleIntervals(visibles, 0, math.MaxInt64),
lookupFileId: nil,
readerLock: sync.Mutex{},
fileSize: 10,
chunkCache: &mockChunkCache{},
readerCache: newReaderCache(3, &mockChunkCache{}, nil),
readerPattern: NewReaderPattern(),
}
@ -116,10 +119,9 @@ func TestReaderAt0(t *testing.T) {
readerAt := &ChunkReadAt{
chunkViews: ViewFromVisibleIntervals(visibles, 0, math.MaxInt64),
lookupFileId: nil,
readerLock: sync.Mutex{},
fileSize: 10,
chunkCache: &mockChunkCache{},
readerCache: newReaderCache(3, &mockChunkCache{}, nil),
readerPattern: NewReaderPattern(),
}
@ -145,10 +147,9 @@ func TestReaderAt1(t *testing.T) {
readerAt := &ChunkReadAt{
chunkViews: ViewFromVisibleIntervals(visibles, 0, math.MaxInt64),
lookupFileId: nil,
readerLock: sync.Mutex{},
fileSize: 20,
chunkCache: &mockChunkCache{},
readerCache: newReaderCache(3, &mockChunkCache{}, nil),
readerPattern: NewReaderPattern(),
}

187
weed/filer/reader_cache.go

@ -0,0 +1,187 @@
package filer
import (
"fmt"
"github.com/chrislusf/seaweedfs/weed/glog"
"github.com/chrislusf/seaweedfs/weed/util/chunk_cache"
"github.com/chrislusf/seaweedfs/weed/util/mem"
"github.com/chrislusf/seaweedfs/weed/wdclient"
"sync"
"time"
)
type ReaderCache struct {
chunkCache chunk_cache.ChunkCache
lookupFileIdFn wdclient.LookupFileIdFunctionType
sync.Mutex
downloaders map[string]*SingleChunkCacher
limit int
}
type SingleChunkCacher struct {
sync.RWMutex
parent *ReaderCache
chunkFileId string
data []byte
err error
cipherKey []byte
isGzipped bool
chunkSize int
shouldCache bool
wg sync.WaitGroup
completedTime time.Time
}
func newReaderCache(limit int, chunkCache chunk_cache.ChunkCache, lookupFileIdFn wdclient.LookupFileIdFunctionType) *ReaderCache {
return &ReaderCache{
limit: limit,
chunkCache: chunkCache,
lookupFileIdFn: lookupFileIdFn,
downloaders: make(map[string]*SingleChunkCacher),
}
}
func (rc *ReaderCache) MaybeCache(fileId string, cipherKey []byte, isGzipped bool, chunkSize int) {
rc.Lock()
defer rc.Unlock()
if _, found := rc.downloaders[fileId]; found {
return
}
if rc.lookupFileIdFn == nil {
return
}
// if too many, delete one of them?
glog.V(0).Infof("downloader2 %d", len(rc.downloaders))
if len(rc.downloaders) >= rc.limit {
oldestFid, oldestTime := "", time.Now()
for fid, downloader := range rc.downloaders {
if !downloader.completedTime.IsZero() {
if downloader.completedTime.Before(oldestTime) {
oldestFid, oldestTime = fid, downloader.completedTime
}
}
}
if oldestFid != "" {
oldDownloader := rc.downloaders[oldestFid]
delete(rc.downloaders, oldestFid)
oldDownloader.destroy()
} else {
// if still no slots, return
return
}
}
cacher := newSingleChunkCacher(rc, fileId, cipherKey, isGzipped, chunkSize, false)
cacher.wg.Add(1)
go cacher.startCaching()
cacher.wg.Wait()
rc.downloaders[fileId] = cacher
return
}
func (rc *ReaderCache) ReadChunkAt(buffer []byte, fileId string, cipherKey []byte, isGzipped bool, offset int64, chunkSize int, shouldCache bool) (int, error) {
rc.Lock()
defer rc.Unlock()
if cacher, found := rc.downloaders[fileId]; found {
return cacher.readChunkAt(buffer, offset)
}
if shouldCache || rc.lookupFileIdFn == nil {
n, err := rc.chunkCache.ReadChunkAt(buffer, fileId, uint64(offset))
if n > 0 {
return n, err
}
}
glog.V(0).Infof("downloader1 %d", len(rc.downloaders))
if len(rc.downloaders) >= rc.limit {
oldestFid, oldestTime := "", time.Now()
for fid, downloader := range rc.downloaders {
if !downloader.completedTime.IsZero() {
if downloader.completedTime.Before(oldestTime) {
oldestFid, oldestTime = fid, downloader.completedTime
}
}
}
if oldestFid != "" {
oldDownloader := rc.downloaders[oldestFid]
delete(rc.downloaders, oldestFid)
oldDownloader.destroy()
}
}
cacher := newSingleChunkCacher(rc, fileId, cipherKey, isGzipped, chunkSize, shouldCache)
cacher.wg.Add(1)
go cacher.startCaching()
cacher.wg.Wait()
rc.downloaders[fileId] = cacher
return cacher.readChunkAt(buffer, offset)
}
func (rc *ReaderCache) destroy() {
rc.Lock()
defer rc.Unlock()
for _, downloader := range rc.downloaders {
downloader.destroy()
}
}
func newSingleChunkCacher(parent *ReaderCache, fileId string, cipherKey []byte, isGzipped bool, chunkSize int, shouldCache bool) *SingleChunkCacher {
t := &SingleChunkCacher{
parent: parent,
chunkFileId: fileId,
cipherKey: cipherKey,
isGzipped: isGzipped,
chunkSize: chunkSize,
shouldCache: shouldCache,
}
return t
}
func (s *SingleChunkCacher) startCaching() {
s.Lock()
defer s.Unlock()
s.wg.Done() // means this has been started
urlStrings, err := s.parent.lookupFileIdFn(s.chunkFileId)
if err != nil {
s.err = fmt.Errorf("operation LookupFileId %s failed, err: %v", s.chunkFileId, err)
return
}
s.data = mem.Allocate(s.chunkSize)
_, s.err = retriedFetchChunkData(s.data, urlStrings, s.cipherKey, s.isGzipped, true, 0)
if s.err != nil {
mem.Free(s.data)
s.data = nil
return
}
s.completedTime = time.Now()
if s.shouldCache {
s.parent.chunkCache.SetChunk(s.chunkFileId, s.data)
}
return
}
func (s *SingleChunkCacher) destroy() {
if s.data != nil {
mem.Free(s.data)
s.data = nil
}
}
func (s *SingleChunkCacher) readChunkAt(buf []byte, offset int64) (int, error) {
s.RLock()
defer s.RUnlock()
return copy(buf, s.data[offset:]), s.err
}

18
weed/filer/stream.go

@ -133,30 +133,30 @@ func writeZero(w io.Writer, size int64) (err error) {
return
}
func ReadAll(masterClient *wdclient.MasterClient, chunks []*filer_pb.FileChunk) ([]byte, error) {
buffer := bytes.Buffer{}
func ReadAll(buffer []byte, masterClient *wdclient.MasterClient, chunks []*filer_pb.FileChunk) error {
lookupFileIdFn := func(fileId string) (targetUrls []string, err error) {
return masterClient.LookupFileId(fileId)
}
chunkViews := ViewFromChunks(lookupFileIdFn, chunks, 0, math.MaxInt64)
chunkViews := ViewFromChunks(lookupFileIdFn, chunks, 0, int64(len(buffer)))
idx := 0
for _, chunkView := range chunkViews {
urlStrings, err := lookupFileIdFn(chunkView.FileId)
if err != nil {
glog.V(1).Infof("operation LookupFileId %s failed, err: %v", chunkView.FileId, err)
return nil, err
return err
}
data, err := retriedFetchChunkData(urlStrings, chunkView.CipherKey, chunkView.IsGzipped, chunkView.IsFullChunk(), chunkView.Offset, int(chunkView.Size))
n, err := retriedFetchChunkData(buffer[idx:idx+int(chunkView.Size)], urlStrings, chunkView.CipherKey, chunkView.IsGzipped, chunkView.IsFullChunk(), chunkView.Offset)
if err != nil {
return nil, err
return err
}
buffer.Write(data)
idx += n
}
return buffer.Bytes(), nil
return nil
}
// ---------------- ChunkStreamReader ----------------------------------

5
weed/server/filer_server_handlers_read.go

@ -4,6 +4,7 @@ import (
"bytes"
"context"
"fmt"
"github.com/chrislusf/seaweedfs/weed/util/mem"
"io"
"mime"
"net/http"
@ -189,7 +190,9 @@ func (fs *FilerServer) GetOrHeadHandler(w http.ResponseWriter, r *http.Request)
}
width, height, mode, shouldResize := shouldResizeImages(ext, r)
if shouldResize {
data, err := filer.ReadAll(fs.filer.MasterClient, entry.Chunks)
data := mem.Allocate(int(totalSize))
defer mem.Free(data)
err := filer.ReadAll(data, fs.filer.MasterClient, entry.Chunks)
if err != nil {
glog.Errorf("failed to read %s: %v", path, err)
w.WriteHeader(http.StatusNotModified)

103
weed/util/chunk_cache/chunk_cache.go

@ -11,8 +11,6 @@ import (
var ErrorOutOfBounds = errors.New("attempt to read out of bounds")
type ChunkCache interface {
GetChunk(fileId string, minSize uint64) (data []byte)
GetChunkSlice(fileId string, offset, length uint64) []byte
ReadChunkAt(data []byte, fileId string, offset uint64) (n int, err error)
SetChunk(fileId string, data []byte)
}
@ -45,107 +43,6 @@ func NewTieredChunkCache(maxEntries int64, dir string, diskSizeInUnit int64, uni
return c
}
func (c *TieredChunkCache) GetChunk(fileId string, minSize uint64) (data []byte) {
if c == nil {
return
}
c.RLock()
defer c.RUnlock()
return c.doGetChunk(fileId, minSize)
}
func (c *TieredChunkCache) doGetChunk(fileId string, minSize uint64) (data []byte) {
if minSize <= c.onDiskCacheSizeLimit0 {
data = c.memCache.GetChunk(fileId)
if len(data) >= int(minSize) {
return data
}
}
fid, err := needle.ParseFileIdFromString(fileId)
if err != nil {
glog.Errorf("failed to parse file id %s", fileId)
return nil
}
if minSize <= c.onDiskCacheSizeLimit0 {
data = c.diskCaches[0].getChunk(fid.Key)
if len(data) >= int(minSize) {
return data
}
}
if minSize <= c.onDiskCacheSizeLimit1 {
data = c.diskCaches[1].getChunk(fid.Key)
if len(data) >= int(minSize) {
return data
}
}
{
data = c.diskCaches[2].getChunk(fid.Key)
if len(data) >= int(minSize) {
return data
}
}
return nil
}
func (c *TieredChunkCache) GetChunkSlice(fileId string, offset, length uint64) []byte {
if c == nil {
return nil
}
c.RLock()
defer c.RUnlock()
return c.doGetChunkSlice(fileId, offset, length)
}
func (c *TieredChunkCache) doGetChunkSlice(fileId string, offset, length uint64) (data []byte) {
minSize := offset + length
if minSize <= c.onDiskCacheSizeLimit0 {
data, err := c.memCache.getChunkSlice(fileId, offset, length)
if err != nil {
glog.Errorf("failed to read from memcache: %s", err)
}
if len(data) >= int(minSize) {
return data
}
}
fid, err := needle.ParseFileIdFromString(fileId)
if err != nil {
glog.Errorf("failed to parse file id %s", fileId)
return nil
}
if minSize <= c.onDiskCacheSizeLimit0 {
data = c.diskCaches[0].getChunkSlice(fid.Key, offset, length)
if len(data) >= int(minSize) {
return data
}
}
if minSize <= c.onDiskCacheSizeLimit1 {
data = c.diskCaches[1].getChunkSlice(fid.Key, offset, length)
if len(data) >= int(minSize) {
return data
}
}
{
data = c.diskCaches[2].getChunkSlice(fid.Key, offset, length)
if len(data) >= int(minSize) {
return data
}
}
return nil
}
func (c *TieredChunkCache) ReadChunkAt(data []byte, fileId string, offset uint64) (n int, err error) {
if c == nil {
return 0, nil

Loading…
Cancel
Save