Files
seaweedfs/weed/s3api/filer_util.go
T
Chris LuandGitHub 7bb0a1c127 s3: replay a delete whose reply the transport dropped (#11022)
* s3: stop retrying a delete the filer refused for a non-empty folder

The filer looked and the children are there, so the answer will not change.
retryFilerOp spent six attempts and up to 3.1s of backoff on it before the
caller could act on the condition it was already holding.

Claude-Session: https://claude.ai/code/session_01XqaJrwgXQ5GSUpyzRbe5nD

* s3: thread the request context through the unversioned delete path

doDeleteEntry issued every DeleteEntry on context.Background(), so an S3
client that hung up left the gateway working on its behalf, out of reach of
both cancellation and the per-request retry allowance that
DeleteMultipleObjectsHandler installs.

Claude-Session: https://claude.ai/code/session_01XqaJrwgXQ5GSUpyzRbe5nD

* s3: treat a cancelled filer RPC as terminal, not transient

isRetryableFilerErr matched context.Canceled and DeadlineExceeded by
sentinel, which only holds while the error is still local. Once it has
crossed gRPC it is a status, so an abandoned request was retried six times
on behalf of a caller that had already gone.

Claude-Session: https://claude.ai/code/session_01XqaJrwgXQ5GSUpyzRbe5nD

* s3: replay a delete whose reply the transport dropped

A delete is idempotent at the filer, which answers an entry that is already
gone with an empty resp.Error, so a reply lost in transit can be reissued
rather than surfaced. Surfaced, it becomes a 500 on the bucket delete, which
boto3 resends and is then answered NoSuchBucket, or a per-key InternalError
inside the 200 of a multi-object delete, which no SDK retries at all.

The replay runs through retryFilerOp, so it draws on the allowance the
request already installs rather than paying a backoff per key, and stops for
a caller that has gone. rm and rmObject re-enter WithFilerClient per attempt,
so each one walks the failover list again on a connection the failed attempt
had invalidated; the multi-object loop holds one client for the batch, so
there the replay reuses it.

Classification stays structural. The filer reports its own refusals in
resp.Error, which carries no status and has the deleted path - and, for a
recursive delete, the children it stopped on - formatted into it, so no key
name can steer the decision either way.

rm and rmObject now take the caller's context. Cleanup and rollback paths
pass context.Background() deliberately: they have to run whether or not the
caller is still waiting.

Claude-Session: https://claude.ai/code/session_01XqaJrwgXQ5GSUpyzRbe5nD

* s3: share one retry allowance across multipart completion cleanup

The unused-entry loop deletes once per entry, and each delete now retries,
so a filer that stays unavailable held the response for 3.1s per entry after
the object was already committed.

Claude-Session: https://claude.ai/code/session_01XqaJrwgXQ5GSUpyzRbe5nD
2026-08-28 14:30:21 -07:00

271 lines
8.9 KiB
Go

package s3api
import (
"context"
"errors"
"fmt"
"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/util"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/status"
)
func (s3a *S3ApiServer) mkdir(parentDirectoryPath string, dirName string, fn func(entry *filer_pb.Entry)) error {
return filer_pb.Mkdir(context.Background(), s3a, parentDirectoryPath, dirName, fn)
}
func (s3a *S3ApiServer) mkFile(parentDirectoryPath string, fileName string, chunks []*filer_pb.FileChunk, fn func(entry *filer_pb.Entry)) error {
err := filer_pb.MkFile(context.Background(), s3a, parentDirectoryPath, fileName, chunks, fn)
if errors.Is(err, filer_pb.ErrExistingIsDirectory) && !isReservedDirectoryName(fileName) {
// Other keys are nested under this one, so the object goes onto the directory
// they live in - the same place a PutObject of this key writes it.
err = filer_pb.MkFile(context.Background(), s3a, parentDirectoryPath, fileName, chunks, func(entry *filer_pb.Entry) {
if fn != nil {
fn(entry)
}
entry.MarkPrefixObject()
})
}
return err
}
func (s3a *S3ApiServer) list(parentDirectoryPath, prefix, startFrom string, inclusive bool, limit uint32) (entries []*filer_pb.Entry, isLast bool, err error) {
return listWithRetry(parentDirectoryPath, func() (entries []*filer_pb.Entry, isLast bool, err error) {
err = filer_pb.List(context.Background(), s3a, parentDirectoryPath, prefix, func(entry *filer_pb.Entry, isLastEntry bool) error {
entries = append(entries, entry)
if isLastEntry {
isLast = true
}
return nil
}, startFrom, inclusive, limit)
if len(entries) == 0 {
isLast = true
}
return
})
}
// A listing has no side effects and collects into a fresh slice per attempt, so
// a replay can neither duplicate nor drop entries; the bound caps a filer that
// is genuinely down at two extra attempts and 300ms of added wait.
const (
listRetryAttempts = 3
listRetryInitialBackoff = 100 * time.Millisecond
)
// isRetryableListError defers to util.IsTransientError, which reads the status
// the filer sent rather than the message DoSeaweedListWithSnapshot builds around
// it out of the bucket and prefix the client chose. Not-found is authoritative
// and must reach the caller unchanged.
func isRetryableListError(err error) bool {
return err != nil && !isFilerNotFound(err) && util.IsTransientError(err)
}
// listWithRetry replays doList while the filer answers with a transient error.
// Both failure points, the ListEntries call itself and the stream.Recv that
// follows it, surface as a plain error out of filer_pb.List, so a single retry
// point above it covers both.
func listWithRetry(parentDirectoryPath string, doList func() (entries []*filer_pb.Entry, isLast bool, err error)) (entries []*filer_pb.Entry, isLast bool, err error) {
backoff := listRetryInitialBackoff
for attempt := 1; ; attempt++ {
entries, isLast, err = doList()
if err == nil || attempt >= listRetryAttempts || !isRetryableListError(err) {
return entries, isLast, err
}
glog.V(1).Infof("list %s attempt %d/%d hit a transient error, retrying in %v: %v", parentDirectoryPath, attempt, listRetryAttempts, backoff, err)
time.Sleep(backoff)
backoff *= 2
}
}
// A delete is idempotent at the filer, which answers an entry that is already
// gone with an empty resp.Error, so a reply the transport dropped can be
// reissued instead of surfaced: as a 500 on the bucket delete, or as a per-key
// InternalError inside the 200 of a multi-object delete, which no SDK retries.
// Each attempt re-enters WithFilerClient, so it walks the failover list again
// on a connection the failed one had invalidated.
func (s3a *S3ApiServer) rm(ctx context.Context, parentDirectoryPath, entryName string, isDeleteData, isRecursive bool) error {
return retryFilerOp(ctx, "rm "+parentDirectoryPath+"/"+entryName, func() error {
return s3a.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
return doDeleteEntry(ctx, client, parentDirectoryPath, entryName, isDeleteData, isRecursive)
})
})
}
func (s3a *S3ApiServer) rmObject(ctx context.Context, parentDirectoryPath, entryName string, isDeleteData, isRecursive bool) error {
return retryFilerOp(ctx, "rmObject "+parentDirectoryPath+"/"+entryName, func() error {
return s3a.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
return deleteObjectEntry(ctx, client, parentDirectoryPath, entryName, isDeleteData, isRecursive)
})
})
}
func deleteObjectEntry(ctx context.Context, client filer_pb.SeaweedFilerClient, parentDirectoryPath, entryName string, isDeleteData, isRecursive bool) error {
err := doDeleteEntry(ctx, client, parentDirectoryPath, entryName, isDeleteData, isRecursive)
if err == nil {
return nil
}
if !errors.Is(err, filer.ErrNonEmptyFolder) {
return err
}
return demoteDirectoryMarkerToImplicitDirectory(ctx, client, parentDirectoryPath, entryName)
}
func doDeleteEntry(ctx context.Context, client filer_pb.SeaweedFilerClient, parentDirectoryPath string, entryName string, isDeleteData bool, isRecursive bool) error {
request := &filer_pb.DeleteEntryRequest{
Directory: parentDirectoryPath,
Name: entryName,
IsDeleteData: isDeleteData,
IsRecursive: isRecursive,
IgnoreRecursiveError: true,
}
glog.V(1).Infof("delete entry %v/%v: %v", parentDirectoryPath, entryName, request)
if resp, err := client.DeleteEntry(ctx, request); err != nil {
glog.V(1).Infof("delete entry %v: %v", request, err)
return fmt.Errorf("delete entry %s/%s: %w", parentDirectoryPath, entryName, err)
} else {
if resp.Error != "" {
// the path wrapped in here is the client's, so classify the filer's
// text now, while it still stands alone
return fmt.Errorf("delete entry %s/%s: %w", parentDirectoryPath, entryName, filer.DeleteEntryError(resp.Error))
}
}
return nil
}
func demoteDirectoryMarkerToImplicitDirectory(ctx context.Context, client filer_pb.SeaweedFilerClient, parentDirectoryPath, entryName string) error {
resp, err := filer_pb.LookupEntry(ctx, client, &filer_pb.LookupDirectoryEntryRequest{
Directory: parentDirectoryPath,
Name: entryName,
})
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return nil
}
return fmt.Errorf("lookup entry %s/%s: %w", parentDirectoryPath, entryName, err)
}
if resp.Entry == nil || !resp.Entry.IsDirectory {
return nil
}
if !resp.Entry.IsDirectoryKeyObject() {
return nil
}
clearDirectoryMarkerMetadata(resp.Entry)
if err := filer_pb.UpdateEntry(ctx, client, &filer_pb.UpdateEntryRequest{
Directory: parentDirectoryPath,
Entry: resp.Entry,
}); err != nil {
if errors.Is(err, filer_pb.ErrNotFound) || status.Code(err) == codes.NotFound {
return nil
}
return fmt.Errorf("update entry %s/%s: %w", parentDirectoryPath, entryName, err)
}
return nil
}
func clearDirectoryMarkerMetadata(entry *filer_pb.Entry) {
if entry == nil {
return
}
if entry.Attributes == nil {
entry.Attributes = &filer_pb.FuseAttributes{}
}
entry.Attributes.Mime = ""
entry.Attributes.Md5 = nil
entry.Attributes.FileSize = 0
entry.Content = nil
entry.Chunks = nil
if len(entry.Extended) == 0 {
return
}
filtered := make(map[string][]byte)
for k, v := range entry.Extended {
lowerKey := strings.ToLower(k)
if lowerKey == s3_constants.SeaweedFSPrefixObject {
// The path is a plain directory again, not a key of its own.
continue
}
if strings.HasPrefix(lowerKey, "xattr-") || strings.HasPrefix(lowerKey, s3_constants.SeaweedFSInternalPrefix) {
filtered[k] = v
}
}
if len(filtered) == 0 {
entry.Extended = nil
return
}
entry.Extended = filtered
}
func (s3a *S3ApiServer) exists(parentDirectoryPath string, entryName string, isDirectory bool) (exists bool, err error) {
return filer_pb.Exists(context.Background(), s3a, parentDirectoryPath, entryName, isDirectory)
}
func (s3a *S3ApiServer) getEntry(parentDirectoryPath, entryName string) (entry *filer_pb.Entry, err error) {
fullPath := util.NewFullPath(parentDirectoryPath, entryName)
entry, _, _, err = filer_pb.GetEntry(context.Background(), s3a, fullPath)
return entry, err
}
func (s3a *S3ApiServer) updateEntry(parentDirectoryPath string, newEntry *filer_pb.Entry) error {
updateEntryRequest := &filer_pb.UpdateEntryRequest{
Directory: parentDirectoryPath,
Entry: newEntry,
}
err := s3a.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
err := filer_pb.UpdateEntry(context.Background(), client, updateEntryRequest)
if err != nil {
return err
}
return nil
})
return err
}
func (s3a *S3ApiServer) getCollectionName(bucket string) string {
if s3a.option.FilerGroup != "" {
return fmt.Sprintf("%s_%s", s3a.option.FilerGroup, bucket)
}
return bucket
}
func objectKey(key *string) *string {
if strings.HasPrefix(*key, "/") {
t := (*key)[1:]
return &t
}
return key
}