Files
seaweedfs/weed/filer/filerstore_wrapper.go
T
Peter Dodd 5f1a742726 fix: return NoSuchKey and drop stale entries when a remote-mounted object is gone from the remote (#11680)
A remote-only entry whose object was deleted from the remote storage
outside the filer answered GET with 500 and stayed in the filer. The
remote's not-found was lost on the way: the backends' ReadFile returned
it as an untyped error, so FetchAndWriteNeedle failed with codes.Unknown
and nothing downstream could tell it from any other failure.

- remote_storage: GCS, S3 (NoSuchKey) and Azure (BlobNotFound) reads
  return ErrRemoteObjectNotFound. GCS reports a missing bucket the same
  way as a missing object, so it confirms the bucket with a listing.
- volume server and filer: the not-found crosses gRPC as codes.NotFound
  carrying the sentinel's text, and the filer's cache RPC returns
  codes.NotFound, which the S3 gateway already maps to NoSuchKey.
- s3api: the origin fallback answers NoSuchKey on a confirmed not-found.
- filer: a confirmed not-found removes the stale entry, so the lazy
  remote-metadata cache converges on the remote. Only remote-only files
  outside .versions and without an active object lock are removed, only
  if unchanged since the fetch (checked on the object's write owner,
  under the lock S3 object writes take), and with a metadata-only
  delete: the filer skips its inline remote delete, and the delete
  events' entries carry a marker that makes filer.remote.sync and
  filer.remote.gateway skip their remote delete, while filer.sync still
  replicates it. The replicated DeleteEntryRequest carries
  keep_remote_object, so the destination's delete events are marked too.
  The store drops the marker from every write, so clients cannot plant
  it.
2026-10-10 11:00:04 +08:00

468 lines
15 KiB
Go

package filer
import (
"context"
"fmt"
"io"
"math"
"strings"
"time"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/viant/ptrie"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/stats"
"github.com/seaweedfs/seaweedfs/weed/util"
)
var (
_ = VirtualFilerStore(&FilerStoreWrapper{})
_ = Debuggable(&FilerStoreWrapper{})
)
type VirtualFilerStore interface {
FilerStore
DeleteHardLink(ctx context.Context, hardLinkId HardLinkId) error
DeleteOneEntry(ctx context.Context, entry *Entry) error
InsertEntryKnownAbsent(ctx context.Context, entry *Entry) error
AddPathSpecificStore(path string, storeId string, store FilerStore)
OnBucketCreation(bucket string)
OnBucketDeletion(bucket string)
CanDropWholeBucket() bool
SameActualStore(a, b util.FullPath) bool
}
type FilerStoreWrapper struct {
defaultStore FilerStore
pathToStore ptrie.Trie[string]
storeIdToStore map[string]FilerStore
hasPathSpecificStore bool // fast check to skip MatchPrefix when no path-specific stores
}
func NewFilerStoreWrapper(store FilerStore) *FilerStoreWrapper {
if innerStore, ok := store.(*FilerStoreWrapper); ok {
return innerStore
}
return &FilerStoreWrapper{
defaultStore: store,
pathToStore: ptrie.New[string](),
storeIdToStore: make(map[string]FilerStore),
}
}
func (fsw *FilerStoreWrapper) CanDropWholeBucket() bool {
if ba, ok := fsw.defaultStore.(BucketAware); ok {
return ba.CanDropWholeBucket()
}
return false
}
func (fsw *FilerStoreWrapper) OnBucketCreation(bucket string) {
for _, store := range fsw.storeIdToStore {
if ba, ok := store.(BucketAware); ok {
ba.OnBucketCreation(bucket)
}
}
if ba, ok := fsw.defaultStore.(BucketAware); ok {
ba.OnBucketCreation(bucket)
}
}
func (fsw *FilerStoreWrapper) OnBucketDeletion(bucket string) {
for _, store := range fsw.storeIdToStore {
if ba, ok := store.(BucketAware); ok {
ba.OnBucketDeletion(bucket)
}
}
if ba, ok := fsw.defaultStore.(BucketAware); ok {
ba.OnBucketDeletion(bucket)
}
}
func (fsw *FilerStoreWrapper) AddPathSpecificStore(path string, storeId string, store FilerStore) {
fsw.storeIdToStore[storeId] = NewFilerStorePathTranslator(path, store)
err := fsw.pathToStore.Put([]byte(path), storeId)
if err != nil {
glog.Fatalf("put path specific store: %v", err)
}
fsw.hasPathSpecificStore = true
}
func (fsw *FilerStoreWrapper) getActualStore(path util.FullPath) (store FilerStore) {
store = fsw.defaultStore
// Fast path: skip MatchPrefix if no path-specific stores are configured (common case)
if !fsw.hasPathSpecificStore {
return
}
if path == "/" || path == "//" {
return
}
var storeId string
fsw.pathToStore.MatchPrefix([]byte(path), func(key []byte, value string) bool {
storeId = value
return false
})
if storeId != "" {
store = fsw.storeIdToStore[storeId]
}
return
}
// SameActualStore reports whether two paths resolve to the same underlying
// store. When path-specific stores are configured, different subtrees may
// be served by different backends.
func (fsw *FilerStoreWrapper) SameActualStore(a, b util.FullPath) bool {
return fsw.getActualStore(a) == fsw.getActualStore(b)
}
func (fsw *FilerStoreWrapper) getDefaultStore() (store FilerStore) {
return fsw.defaultStore
}
func (fsw *FilerStoreWrapper) GetName() string {
return fsw.getDefaultStore().GetName()
}
func (fsw *FilerStoreWrapper) Initialize(configuration util.Configuration, prefix string) error {
return fsw.getDefaultStore().Initialize(configuration, prefix)
}
func (fsw *FilerStoreWrapper) InsertEntry(ctx context.Context, entry *Entry) error {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(entry.FullPath)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "insert").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "insert").Observe(time.Since(start).Seconds())
}()
filer_pb.BeforeEntrySerialization(entry.GetChunks())
normalizeEntryMimeForStore(entry)
delete(entry.Extended, ExtKeepRemoteObjectKey)
if len(entry.HardLinkId) > 0 {
glog.V(4).InfofCtx(ctx, "InsertEntry %s has HardLinkId %x counter=%d",
entry.FullPath, entry.HardLinkId, entry.HardLinkCounter)
}
if err := fsw.handleUpdateToHardLinks(ctx, entry); err != nil {
return err
}
return actualStore.InsertEntry(ctx, entry)
}
// InsertEntryKnownAbsent skips the pre-insert FindEntry path when the caller has
// already established that the target path does not exist.
func (fsw *FilerStoreWrapper) InsertEntryKnownAbsent(ctx context.Context, entry *Entry) error {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(entry.FullPath)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "insert").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "insert").Observe(time.Since(start).Seconds())
}()
filer_pb.BeforeEntrySerialization(entry.GetChunks())
normalizeEntryMimeForStore(entry)
delete(entry.Extended, ExtKeepRemoteObjectKey)
if len(entry.HardLinkId) > 0 {
glog.V(4).InfofCtx(ctx, "InsertEntryKnownAbsent %s has HardLinkId %x counter=%d",
entry.FullPath, entry.HardLinkId, entry.HardLinkCounter)
if err := fsw.setHardLink(ctx, entry); err != nil {
return fmt.Errorf("setHardLink %x: %v", entry.HardLinkId, err)
}
}
return actualStore.InsertEntry(ctx, entry)
}
func (fsw *FilerStoreWrapper) UpdateEntry(ctx context.Context, entry *Entry) error {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(entry.FullPath)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "update").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "update").Observe(time.Since(start).Seconds())
}()
filer_pb.BeforeEntrySerialization(entry.GetChunks())
normalizeEntryMimeForStore(entry)
delete(entry.Extended, ExtKeepRemoteObjectKey)
if len(entry.HardLinkId) > 0 {
glog.V(4).InfofCtx(ctx, "UpdateEntry %s has HardLinkId %x counter=%d",
entry.FullPath, entry.HardLinkId, entry.HardLinkCounter)
}
if err := fsw.handleUpdateToHardLinks(ctx, entry); err != nil {
return err
}
return actualStore.UpdateEntry(ctx, entry)
}
func normalizeEntryMimeForStore(entry *Entry) {
if entry == nil {
return
}
if !entry.IsDirectory() && entry.Mime == "application/octet-stream" {
entry.Mime = ""
}
}
func (fsw *FilerStoreWrapper) FindEntry(ctx context.Context, fp util.FullPath) (entry *Entry, err error) {
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(fp)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "find").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "find").Observe(time.Since(start).Seconds())
}()
entry, err = actualStore.FindEntry(ctx, fp)
// glog.V(4).Infof("FindEntry %s: %v", fp, err)
if err != nil {
if fsw.CanDropWholeBucket() && strings.Contains(err.Error(), "Table") && strings.Contains(err.Error(), "doesn't exist") {
err = filer_pb.ErrNotFound
}
return nil, err
}
fsw.maybeReadHardLink(ctx, entry)
filer_pb.AfterEntryDeserialization(entry.GetChunks())
return
}
func (fsw *FilerStoreWrapper) DeleteEntry(ctx context.Context, fp util.FullPath) (err error) {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(fp)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "delete").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "delete").Observe(time.Since(start).Seconds())
}()
existingEntry, findErr := fsw.FindEntry(ctx, fp)
if findErr == filer_pb.ErrNotFound || existingEntry == nil {
return nil
}
if len(existingEntry.HardLinkId) != 0 {
// remove hard link
op := ctx.Value("OP")
if op != "MV" {
glog.V(4).InfofCtx(ctx, "DeleteHardLink %s", existingEntry.FullPath)
if hlErr := fsw.DeleteHardLink(ctx, existingEntry.HardLinkId); hlErr != nil {
// Log but continue: the directory entry must be removed
// even if hard link counter cleanup fails, otherwise the
// parent directory cannot be removed (rmdir ENOTEMPTY).
glog.Warningf("DeleteHardLink %s (id %x): %v", existingEntry.FullPath, existingEntry.HardLinkId, hlErr)
}
}
}
return actualStore.DeleteEntry(ctx, fp)
}
func (fsw *FilerStoreWrapper) DeleteOneEntry(ctx context.Context, existingEntry *Entry) (err error) {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(existingEntry.FullPath)
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "delete").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "delete").Observe(time.Since(start).Seconds())
}()
if len(existingEntry.HardLinkId) != 0 {
// remove hard link
op := ctx.Value("OP")
if op != "MV" {
glog.V(4).InfofCtx(ctx, "DeleteHardLink %s", existingEntry.FullPath)
if hlErr := fsw.DeleteHardLink(ctx, existingEntry.HardLinkId); hlErr != nil {
// Log the hard link cleanup error but continue to delete
// the directory entry. If we return early here, the entry
// remains in the store and the parent directory cannot be
// removed (rmdir returns ENOTEMPTY).
glog.Warningf("DeleteHardLink %s (id %x): %v", existingEntry.FullPath, existingEntry.HardLinkId, hlErr)
}
}
}
return actualStore.DeleteEntry(ctx, existingEntry.FullPath)
}
func (fsw *FilerStoreWrapper) DeleteFolderChildren(ctx context.Context, fp util.FullPath) (err error) {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(fp + "/")
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "deleteFolderChildren").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "deleteFolderChildren").Observe(time.Since(start).Seconds())
}()
return actualStore.DeleteFolderChildren(ctx, fp)
}
func (fsw *FilerStoreWrapper) ListDirectoryEntries(ctx context.Context, dirPath util.FullPath, startFileName string, includeStartFile bool, limit int64, eachEntryFunc ListEachEntryFunc) (string, error) {
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(dirPath + "/")
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "list").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "list").Observe(time.Since(start).Seconds())
}()
// glog.V(4).Infof("ListDirectoryEntries %s from %s limit %d", dirPath, startFileName, limit)
return actualStore.ListDirectoryEntries(ctx, dirPath, startFileName, includeStartFile, limit, func(entry *Entry) (bool, error) {
fsw.maybeReadHardLink(ctx, entry)
filer_pb.AfterEntryDeserialization(entry.GetChunks())
return eachEntryFunc(entry)
})
}
func (fsw *FilerStoreWrapper) ListDirectoryPrefixedEntries(ctx context.Context, dirPath util.FullPath, startFileName string, includeStartFile bool, limit int64, prefix string, eachEntryFunc ListEachEntryFunc) (lastFileName string, err error) {
ctx = context.WithoutCancel(ctx)
actualStore := fsw.getActualStore(dirPath + "/")
stats.FilerStoreCounter.WithLabelValues(actualStore.GetName(), "prefixList").Inc()
start := time.Now()
defer func() {
stats.FilerStoreHistogram.WithLabelValues(actualStore.GetName(), "prefixList").Observe(time.Since(start).Seconds())
}()
if limit > math.MaxInt32-1 {
limit = math.MaxInt32 - 1
}
// glog.V(4).Infof("ListDirectoryPrefixedEntries %s from %s prefix %s limit %d", dirPath, startFileName, prefix, limit)
adjustedEntryFunc := func(entry *Entry) (bool, error) {
fsw.maybeReadHardLink(ctx, entry)
filer_pb.AfterEntryDeserialization(entry.GetChunks())
return eachEntryFunc(entry)
}
lastFileName, err = actualStore.ListDirectoryPrefixedEntries(ctx, dirPath, startFileName, includeStartFile, limit, prefix, adjustedEntryFunc)
if err == ErrUnsupportedListDirectoryPrefixed {
lastFileName, err = fsw.prefixFilterEntries(ctx, dirPath, startFileName, includeStartFile, limit, prefix, adjustedEntryFunc)
}
return lastFileName, err
}
func (fsw *FilerStoreWrapper) prefixFilterEntries(ctx context.Context, dirPath util.FullPath, startFileName string, includeStartFile bool, limit int64, prefix string, eachEntryFunc ListEachEntryFunc) (lastFileName string, err error) {
actualStore := fsw.getActualStore(dirPath + "/")
if prefix == "" {
return actualStore.ListDirectoryEntries(ctx, dirPath, startFileName, includeStartFile, limit, eachEntryFunc)
}
var notPrefixed []*Entry
lastFileName, err = actualStore.ListDirectoryEntries(ctx, dirPath, startFileName, includeStartFile, limit, func(entry *Entry) (bool, error) {
notPrefixed = append(notPrefixed, entry)
return true, nil
})
if err != nil {
return
}
count := int64(0)
for count < limit && len(notPrefixed) > 0 {
for _, entry := range notPrefixed {
if strings.HasPrefix(entry.Name(), prefix) {
count++
res, resErr := eachEntryFunc(entry)
if resErr != nil {
err = fmt.Errorf("failed to process eachEntryFunc for entry %q: %w", entry.Name(), resErr)
return
}
if !res {
return
}
if count >= limit {
break
}
}
}
if count < limit && lastFileName < prefix {
notPrefixed = notPrefixed[:0]
lastFileName, err = actualStore.ListDirectoryEntries(ctx, dirPath, lastFileName, false, limit, func(entry *Entry) (bool, error) {
notPrefixed = append(notPrefixed, entry)
return true, nil
})
if err != nil {
return
}
} else {
break
}
}
return
}
func (fsw *FilerStoreWrapper) BeginTransaction(ctx context.Context) (context.Context, error) {
if err := ctx.Err(); err != nil {
return nil, err
}
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().BeginTransaction(ctx)
}
func (fsw *FilerStoreWrapper) CommitTransaction(ctx context.Context) error {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().CommitTransaction(ctx)
}
func (fsw *FilerStoreWrapper) RollbackTransaction(ctx context.Context) error {
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().RollbackTransaction(ctx)
}
func (fsw *FilerStoreWrapper) Shutdown() {
fsw.getDefaultStore().Shutdown()
}
func (fsw *FilerStoreWrapper) KvPut(ctx context.Context, key []byte, value []byte) (err error) {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().KvPut(ctx, key, value)
}
func (fsw *FilerStoreWrapper) KvGet(ctx context.Context, key []byte) (value []byte, err error) {
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().KvGet(ctx, key)
}
func (fsw *FilerStoreWrapper) KvDelete(ctx context.Context, key []byte) (err error) {
if err := ctx.Err(); err != nil {
return err
}
ctx = context.WithoutCancel(ctx)
return fsw.getDefaultStore().KvDelete(ctx, key)
}
func (fsw *FilerStoreWrapper) Debug(writer io.Writer) {
if debuggable, ok := fsw.getDefaultStore().(Debuggable); ok {
debuggable.Debug(writer)
}
}