Files
seaweedfs/weed/s3api/s3api_directory_marker.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

113 lines
5.4 KiB
Go

package s3api
import (
"errors"
"net/http"
"strings"
"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/s3api/s3err"
"github.com/seaweedfs/seaweedfs/weed/util"
)
// An explicit directory marker - the key "dir/" created by PutObject on a
// trailing-slash key - is stored as the filer directory itself rather than as an
// object beside it. That makes it a poor fit for versioning: a delete marker would
// have to replace an entry that other keys live under, and the history would have to
// sit inside the directory it describes, where a listing keeps meeting it.
//
// So the key is not versioned. Deleting it does what deleting it in an unversioned
// bucket already does: the directory is removed when nothing is left under it, and
// demoted to a plain directory when children remain. Listings need no version lookup
// to tell what a directory stands for, and a bucket made of directory markers costs
// the same to list versioned as unversioned.
// deleteDirectoryMarker removes the key "<dir>/". Callers hold the object write lock,
// so the entry this decides about cannot change between the read and the delete.
func (s3a *S3ApiServer) deleteDirectoryMarker(r *http.Request, bucket, object string) s3err.ErrorCode {
governanceBypassAllowed := s3a.evaluateGovernanceBypassRequest(r, bucket, object)
markerDir := s3a.bucketDir(bucket) + "/" + strings.TrimSuffix(strings.TrimPrefix(object, "/"), "/")
dir, name := util.FullPath(markerDir).DirAndName()
entry, err := s3a.getEntry(dir, name)
switch {
case errors.Is(err, filer_pb.ErrNotFound):
return s3err.ErrNone // deleting a key that is not there is a success
case err != nil:
// The entry may be a file a child write promoted to a directory, whose data
// belongs to the key without the trailing slash. Deleting without knowing
// would destroy it, so fail and leave the retry to the client.
glog.Errorf("deleteDirectoryMarker: cannot read %s/%s: %v", bucket, object, err)
return s3err.ErrInternalError
case len(entry.GetChunks()) > 0 || entry.IsInRemoteOnly():
// A promoted file, not a marker: "dir/" does not name its data.
glog.V(2).Infof("deleteDirectoryMarker: %s/%s holds uploaded data, leaving it alone", bucket, object)
return s3err.ErrNone
}
// The key is deleted the unversioned way, but Object Lock still covers it: the
// gateway lists it as an object and serves retention set on it. The lock that
// matters is the one on this entry, since that is what is removed -- looking the
// key up instead would answer with a version once the key has a history.
if err := s3a.enforceObjectLockOnEntry(entry, bucket, object, "", governanceBypassAllowed); err != nil {
glog.V(2).Infof("deleteDirectoryMarker: %s/%s is locked: %v", bucket, object, err)
return s3err.ErrAccessDenied
}
// Drop a history an older build recorded for this key. Nothing writes one now, and
// leaving it behind keeps reporting the key in ListObjectVersions, so a history we
// cannot read or remove fails the delete rather than half finishing it.
switch _, historyErr := s3a.getEntry(markerDir, s3_constants.VersionsFolder); {
case historyErr == nil:
// The removal below takes every entry under the key, so each has to be clear
// of a lock of its own.
versionsDir := markerDir + "/" + s3_constants.VersionsFolder
for startFrom := ""; ; {
entries, isLast, listErr := s3a.list(versionsDir, "", startFrom, false, 1000)
if listErr != nil {
glog.Errorf("deleteDirectoryMarker: cannot list history of %s/%s: %v", bucket, object, listErr)
return s3err.ErrInternalError
}
for _, entry := range entries {
startFrom = entry.Name
versionId, named := entry.Extended[s3_constants.ExtVersionIdKey]
if !named {
// An entry an older build left without a version id is what this
// removal is here to clear, but one still under a lock cannot be
// named to check it, so judge it on what it carries itself.
if err := s3a.enforceObjectLockOnEntry(entry, bucket, object, "", governanceBypassAllowed); err != nil {
glog.V(2).Infof("deleteDirectoryMarker: unnamed history entry %s of %s/%s is locked: %v", entry.Name, bucket, object, err)
return s3err.ErrAccessDenied
}
continue
}
if err := s3a.enforceObjectLockProtections(r, bucket, object, string(versionId), governanceBypassAllowed); err != nil {
glog.V(2).Infof("deleteDirectoryMarker: version %s of %s/%s is locked: %v", versionId, bucket, object, err)
return s3err.ErrAccessDenied
}
}
if isLast || len(entries) == 0 {
break
}
}
if rmErr := s3a.rm(r.Context(), markerDir, s3_constants.VersionsFolder, true, true); rmErr != nil {
glog.Errorf("deleteDirectoryMarker: failed to remove stale history of %s/%s: %v", bucket, object, rmErr)
return s3err.ErrInternalError
}
case !errors.Is(historyErr, filer_pb.ErrNotFound):
glog.Errorf("deleteDirectoryMarker: cannot read stale history of %s/%s: %v", bucket, object, historyErr)
return s3err.ErrInternalError
}
if err := s3a.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
return s3a.deleteUnversionedObjectWithClient(r.Context(), client, bucket, object, false)
}); err != nil {
glog.Errorf("deleteDirectoryMarker: failed to delete %s/%s: %v", bucket, object, err)
return s3err.ErrInternalError
}
return s3err.ErrNone
}