Files
seaweedfs/weed/s3api/s3api_internal_lifecycle.go
T
Chris LuandGitHub c6ad6dcf74 feat(s3/lifecycle): sole-survivor delete-marker routing (Phase 5b/2) (#9381)
* feat(s3/lifecycle): sole-survivor delete-marker routing (Phase 5b/2)

Production stores delete markers under <object>.versions/<version-file>;
buildObjectInfo skips every version-folder event because it can't tell
current from noncurrent without sibling state. EXP_DM is the one rule
that can be routed from the file path alone: if the marker is the only
entry under .versions/<key>/ then it's necessarily the latest.

Add SiblingLister; gate the listing on (versioned, version-folder path,
delete-marker entry, EXP_DM rule active for the bucket, lister supplied)
so non-marker version writes pay nothing. The Match carries the LOGICAL
key in ObjectKey and the marker's version_id in VersionID so the
dispatcher can reach deleteSpecificObjectVersion(bucket, logical, vid);
without a version_id the dispatcher would BLOCK and freeze the cursor.

Server-side dispatch re-checks sole-survivor before deleting: another
PUT can land between schedule and dispatch, and identity-CAS only covers
the marker entry, not the directory shape. Lists .versions/<key>/ with
limit 2 and returns NOOP_RESOLVED if count != 1 or the surviving entry
is not the same version.

Fix isDeleteMarkerEntry to compare string(v) == "true" — production
writes []byte("true"), not {1}; every other reader of ExtDeleteMarkerKey
in this repo uses the string predicate.

filerSiblingLister.Count caps at limit 2 — callers only distinguish
"sole survivor" (1) from "more than one" (>=2).

Follows #9373.

* fix(s3/lifecycle): gate sibling listing on active event-driven EXP_DM

hasActionKind tested key presence only; an EXP_DM rule whose action was
inactive (BootstrapComplete=false) or scan-only would still trigger the
sibling-listing RPC even though no match could fire. Replace with a
helper that mirrors the per-key filter Route already applies before
emitting matches: kind matches, snap.Action returns non-nil, IsActive,
and Mode == ModeEventDriven.

Add a regression test that builds a versioned snapshot with the rule
but no PriorStates (action stays inactive) and asserts the lister is
never consulted.

* chore(s3/lifecycle): trim verbose comments

* fix(s3/lifecycle): null version, hard-delete trigger, latest pointer

Three correctness gaps in EXP_DM routing.

1. Null version. SiblingLister.Count saw only .versions/<key>/, but a
   pre-versioning bare-key object survives outside that folder when
   versioning is enabled later. Marker + null would look like count==1
   and EXP_DM would re-expose the old object. Replace Count with
   Survivors which also reports HasNullVersion; route- and dispatch-time
   suppress when set.

2. Hard-delete trigger. The router skipped events with NewEntry==nil so
   a noncurrent hard-delete that left the marker as sole survivor never
   fired EXP_DM. Allow version-folder hard-deletes through; the lister's
   LoneEntry carries the surviving marker's version_id and identity.

3. Latest pointer. checkSoleSurvivorMarker only checked count and
   filename. The worker can race createDeleteMarker between the file
   write and the .versions/ directory metadata update. Require
   ExtLatestVersionIdKey == versionId; missing pointer returns
   retry-later instead of deleting.

Adds a null-version exists check on the dispatch path too.

* fix(s3/lifecycle): normalize null-version lookup errors and detect dir markers

Two correctness bugs in the null-version check.

Route-side: filerSiblingLister called client.LookupDirectoryEntry
directly. SeaweedList wraps not-found via filer_pb.LookupEntry, which
normalizes the gRPC string-mapped not-found into ErrNotFound. The raw
client returns it as a generic error instead, so every absent
null-version (the common case) bubbled up as an error and the router
suppressed every otherwise-valid match. Use filer_pb.LookupEntry.

Both sides: explicit S3 directory-key markers (object names ending in /)
are stored as directory entries with Attributes.Mime set;
processExplicitDirectory in the listing path treats them as null
versions. The previous check was !IsDirectory only, which let the
marker delete proceed and re-expose the directory key. Add
IsDirectoryKeyObject() to both predicates. Also use
util.NewFullPath(...).DirAndName() for the parent/name split so a
trailing-slash key resolves to the same underlying entry path as the
listing code.

* fix(s3/lifecycle): EXP_DM ctx propagation, nil-entry guard, fast-path skip

Three small follow-ups on the EXP_DM dispatch path.

checkSoleSurvivorMarker now takes ctx instead of context.Background()
so worker shutdown / deadlines cancel the SeaweedList RPC instead of
stalling.

If SeaweedList fires the lone callback with entry==nil, firstName
stays empty and the marker-replaced check would short-circuit; that's
the one shape that bypasses the dispatch guard, so retry-later instead.

routeSoleSurvivorMarker now skips the Survivors RPC on regular
non-marker version creates — those always have Count >= 2, so the
listing was wasted load on every versioned write under an EXP_DM rule.
Hard-delete events (NewEntry==nil) and marker creates still flow
through. Added a regression test asserting the regular-create case
doesn't consult the lister.

Documented that logicalKeyFromVersionPath rejects bucket-root markers
intentionally.
2026-05-08 23:24:08 -07:00

352 lines
14 KiB
Go

package s3api
import (
"bytes"
"context"
"errors"
"path"
"strings"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/pb/s3_lifecycle_pb"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3_constants"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3lifecycle"
)
// LifecycleDelete executes one (rule, action) verdict: re-fetch, identity
// CAS, object-lock check, dispatch by kind. Errors surface as outcomes;
// reader cursors and pending state are the worker's concern.
func (s3a *S3ApiServer) LifecycleDelete(ctx context.Context, req *s3_lifecycle_pb.LifecycleDeleteRequest) (*s3_lifecycle_pb.LifecycleDeleteResponse, error) {
if req == nil || req.Bucket == "" || req.ObjectPath == "" {
return blocked("FATAL_EVENT_ERROR: empty bucket or object_path"), nil
}
// MPU init lives at .uploads/<id>/; not handled by getObjectEntry.
if req.ActionKind == s3_lifecycle_pb.ActionKind_ABORT_MPU {
return s3a.lifecycleAbortMPU(ctx, req)
}
entry, err := s3a.getObjectEntry(req.Bucket, req.ObjectPath, req.VersionId)
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) || errors.Is(err, ErrObjectNotFound) || errors.Is(err, ErrVersionNotFound) || errors.Is(err, ErrLatestVersionNotFound) {
return noopResolved("NOT_FOUND"), nil
}
glog.V(1).Infof("lifecycle: live fetch %s/%s@%s: %v", req.Bucket, req.ObjectPath, req.VersionId, err)
return retryLater("TRANSPORT_ERROR: " + err.Error()), nil
}
if !identityMatches(computeEntryIdentity(entry), req.ExpectedIdentity) {
return noopResolved("STALE_IDENTITY"), nil
}
// Lifecycle never bypasses governance/compliance; the http.Request is
// only read when bypass is allowed, so nil is safe here.
if err := s3a.enforceObjectLockProtections(nil, req.Bucket, req.ObjectPath, req.VersionId, false); err != nil {
glog.V(2).Infof("lifecycle: SKIPPED_OBJECT_LOCK %s/%s@%s: %v", req.Bucket, req.ObjectPath, req.VersionId, err)
return &s3_lifecycle_pb.LifecycleDeleteResponse{
Outcome: s3_lifecycle_pb.LifecycleDeleteOutcome_SKIPPED_OBJECT_LOCK,
Reason: err.Error(),
}, nil
}
return s3a.lifecycleDispatch(ctx, req, entry)
}
func (s3a *S3ApiServer) lifecycleDispatch(ctx context.Context, req *s3_lifecycle_pb.LifecycleDeleteRequest, entry *filer_pb.Entry) (*s3_lifecycle_pb.LifecycleDeleteResponse, error) {
switch req.ActionKind {
case s3_lifecycle_pb.ActionKind_EXPIRATION_DAYS, s3_lifecycle_pb.ActionKind_EXPIRATION_DATE:
// Current-version expiration: Enabled -> delete marker; Suspended
// -> delete null + new marker; Off -> remove. Filer errors classify
// as RETRY_LATER; the worker's budget promotes to BLOCKED.
state, vErr := s3a.getVersioningState(req.Bucket)
if vErr != nil {
if errors.Is(vErr, filer_pb.ErrNotFound) {
return noopResolved("BUCKET_NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: versioning lookup: " + vErr.Error()), nil
}
switch state {
case s3_constants.VersioningEnabled:
if _, err := s3a.createDeleteMarker(req.Bucket, req.ObjectPath); err != nil {
return retryLater("TRANSPORT_ERROR: createDeleteMarker: " + err.Error()), nil
}
return done(), nil
case s3_constants.VersioningSuspended:
// Best-effort null delete; NotFound is benign.
if err := s3a.deleteSpecificObjectVersion(req.Bucket, req.ObjectPath, "null"); err != nil {
if !errors.Is(err, filer_pb.ErrNotFound) && !errors.Is(err, ErrVersionNotFound) {
return retryLater("TRANSPORT_ERROR: deleteNullVersion: " + err.Error()), nil
}
}
if _, err := s3a.createDeleteMarker(req.Bucket, req.ObjectPath); err != nil {
return retryLater("TRANSPORT_ERROR: createDeleteMarker: " + err.Error()), nil
}
return done(), nil
default:
err := s3a.WithFilerClient(false, func(c filer_pb.SeaweedFilerClient) error {
return s3a.deleteUnversionedObjectWithClient(c, req.Bucket, req.ObjectPath)
})
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) || errors.Is(err, ErrObjectNotFound) {
return noopResolved("NOT_FOUND_AT_DELETE"), nil
}
return retryLater("TRANSPORT_ERROR: deleteUnversioned: " + err.Error()), nil
}
return done(), nil
}
case s3_lifecycle_pb.ActionKind_NONCURRENT_DAYS,
s3_lifecycle_pb.ActionKind_NEWER_NONCURRENT,
s3_lifecycle_pb.ActionKind_EXPIRED_DELETE_MARKER:
// EXPIRED_DELETE_MARKER targets the marker version itself.
if req.VersionId == "" {
return blocked("FATAL_EVENT_ERROR: version_id required for noncurrent / delete-marker delete"), nil
}
// Latest-pointer guard for noncurrent kinds: refuse to delete
// the version that the .versions/ directory currently points
// to. The router can't always tell current from noncurrent
// without sibling state, so the server checks here.
if req.ActionKind == s3_lifecycle_pb.ActionKind_NONCURRENT_DAYS ||
req.ActionKind == s3_lifecycle_pb.ActionKind_NEWER_NONCURRENT {
isLatest, lookupErr := s3a.isCurrentLatestVersion(req.Bucket, req.ObjectPath, req.VersionId)
if lookupErr != nil {
if errors.Is(lookupErr, filer_pb.ErrNotFound) || errors.Is(lookupErr, ErrObjectNotFound) {
return noopResolved("NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: latest-pointer lookup: " + lookupErr.Error()), nil
}
if isLatest {
return noopResolved("VERSION_IS_LATEST"), nil
}
}
// Re-check sole-survivor: a fresh PUT can land between schedule
// and dispatch. Identity-CAS upstream covers the marker bytes;
// this covers the directory shape.
if req.ActionKind == s3_lifecycle_pb.ActionKind_EXPIRED_DELETE_MARKER {
outcome, err := s3a.checkSoleSurvivorMarker(ctx, req.Bucket, req.ObjectPath, req.VersionId)
if outcome != nil || err != nil {
return outcome, err
}
}
if err := s3a.deleteSpecificObjectVersion(req.Bucket, req.ObjectPath, req.VersionId); err != nil {
if errors.Is(err, filer_pb.ErrNotFound) || errors.Is(err, ErrVersionNotFound) || errors.Is(err, ErrObjectNotFound) {
return noopResolved("NOT_FOUND_AT_DELETE"), nil
}
return retryLater("TRANSPORT_ERROR: deleteSpecificVersion: " + err.Error()), nil
}
return done(), nil
case s3_lifecycle_pb.ActionKind_ABORT_MPU:
return blocked("FATAL_EVENT_ERROR: ABORT_MPU dispatched after fetch"), nil
default:
return blocked("FATAL_EVENT_ERROR: unknown action_kind " + req.ActionKind.String()), nil
}
}
func (s3a *S3ApiServer) lifecycleAbortMPU(ctx context.Context, req *s3_lifecycle_pb.LifecycleDeleteRequest) (*s3_lifecycle_pb.LifecycleDeleteResponse, error) {
// req.ObjectPath is `.uploads/<upload_id>` (set by the router from the
// init directory's bucket-relative path); reject anything that isn't
// exactly that shape so a malformed event can't escalate to a wider rm.
const uploadsPrefix = s3_constants.MultipartUploadsFolder + "/"
if !strings.HasPrefix(req.ObjectPath, uploadsPrefix) {
return blocked("FATAL_EVENT_ERROR: ABORT_MPU object_path missing .uploads/ prefix"), nil
}
uploadID := req.ObjectPath[len(uploadsPrefix):]
// Reject "." and ".." explicitly: util.JoinPath in the filer cleans
// path components, so .uploads/.. would resolve to the bucket root.
if uploadID == "" || uploadID == "." || uploadID == ".." || strings.ContainsRune(uploadID, '/') {
return blocked("FATAL_EVENT_ERROR: ABORT_MPU object_path malformed: " + req.ObjectPath), nil
}
uploadsFolder := s3a.genUploadsFolder(req.Bucket)
// Pre-check existence: filer.DeleteEntry suppresses ErrNotFound and
// returns success, so without this check an already-aborted upload
// would report DONE instead of the correct NOOP_RESOLVED.
exists, err := s3a.exists(uploadsFolder, uploadID, true)
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return noopResolved("NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: exists: " + err.Error()), nil
}
if !exists {
return noopResolved("NOT_FOUND"), nil
}
if err := s3a.rm(uploadsFolder, uploadID, true, true); err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return noopResolved("NOT_FOUND_AT_DELETE"), nil
}
glog.V(1).Infof("lifecycle abort_mpu %s/%s: %v", req.Bucket, req.ObjectPath, err)
return retryLater("TRANSPORT_ERROR: rm: " + err.Error()), nil
}
return done(), nil
}
// checkSoleSurvivorMarker returns nil to proceed with the delete, or a
// terminal response when state has drifted: count != 1, the surviving
// entry is a different version, the .versions/ directory's latest
// pointer doesn't name versionId, or a bare null-version exists outside
// .versions/. Pointer missing while a marker is present is treated as
// retry-later — the create races with the directory metadata update.
func (s3a *S3ApiServer) checkSoleSurvivorMarker(ctx context.Context, bucket, object, versionId string) (*s3_lifecycle_pb.LifecycleDeleteResponse, error) {
bucketDir := s3a.bucketDir(bucket)
versionsDir := bucketDir + "/" + object + s3_constants.VersionsFolder
count := 0
var firstName string
err := s3a.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
return filer_pb.SeaweedList(ctx, client, versionsDir, "", func(entry *filer_pb.Entry, _ bool) error {
count++
if count == 1 && entry != nil {
firstName = entry.Name
}
return nil
}, "", false, 2)
})
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return noopResolved("NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: sole-survivor list: " + err.Error()), nil
}
if count == 0 {
return noopResolved("NOT_FOUND"), nil
}
if count > 1 {
return noopResolved("NOT_SOLE_SURVIVOR"), nil
}
// SeaweedList delivered a single callback but with a nil entry; we
// can't compare names so retry rather than silently bypass the
// marker-replaced check.
if firstName == "" {
return retryLater("PENDING_SURVIVOR_ENTRY"), nil
}
if versionId != "" && firstName != s3a.getVersionFileName(versionId) {
return noopResolved("MARKER_REPLACED"), nil
}
// Latest-pointer check: createDeleteMarker writes the marker file
// and then updates the parent directory's Extended map. Reading
// before the second step lands would see count==1 but no pointer;
// retry-later rather than mistakenly delete.
parent, name := path.Split(versionsDir)
parent = strings.TrimRight(parent, "/")
if parent == "" {
parent = "/"
}
versionsEntry, err := s3a.getEntry(parent, name)
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return noopResolved("NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: latest-pointer lookup: " + err.Error()), nil
}
if versionsEntry == nil {
return noopResolved("NOT_FOUND"), nil
}
latest, hasPointer := versionsEntry.Extended[s3_constants.ExtLatestVersionIdKey]
if !hasPointer || len(latest) == 0 {
return retryLater("PENDING_LATEST_POINTER"), nil
}
if string(latest) != versionId {
return noopResolved("MARKER_NOT_LATEST"), nil
}
// Null-version check: pre-versioning objects survive as the bare
// <bucket>/<key>. Both regular files and explicit S3 directory-key
// markers (object names ending in /) qualify; the listing path
// (s3api_object_versioning.go processExplicitDirectory) treats both
// as the null version. getEntry uses NewFullPath so a trailing slash
// in object splits the same as a regular key.
bareEntry, err := s3a.getEntry(bucketDir, object)
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return nil, nil
}
return retryLater("TRANSPORT_ERROR: null-version lookup: " + err.Error()), nil
}
if bareEntry != nil && (!bareEntry.IsDirectory || bareEntry.IsDirectoryKeyObject()) {
return noopResolved("NULL_VERSION_PRESENT"), nil
}
return nil, nil
}
// isCurrentLatestVersion reports whether versionId is the version the
// .versions/ directory currently points to. SeaweedFS records the latest
// version on the parent directory's Extended map; without consulting it,
// a noncurrent-kind dispatch can't safely distinguish current from
// noncurrent and would risk deleting the live version. Returns
// (false, nil) when the directory has no latest pointer (e.g., the
// bucket isn't versioned in this object's history).
func (s3a *S3ApiServer) isCurrentLatestVersion(bucket, object, versionId string) (bool, error) {
versionsDir := s3a.bucketDir(bucket) + "/" + object + s3_constants.VersionsFolder
parent, name := path.Split(versionsDir)
parent = strings.TrimRight(parent, "/")
if parent == "" {
parent = "/"
}
entry, err := s3a.getEntry(parent, name)
if err != nil {
return false, err
}
if entry == nil || len(entry.Extended) == 0 {
return false, nil
}
latest, ok := entry.Extended[s3_constants.ExtLatestVersionIdKey]
if !ok {
return false, nil
}
return string(latest) == versionId, nil
}
// computeEntryIdentity captures (mtime, size, head fid, sorted-Extended hash):
// an overwrite changes mtime/size/fid; a metadata edit changes Extended; a
// snapshot-restore that preserves mtime+size still differs in head_fid.
func computeEntryIdentity(entry *filer_pb.Entry) *s3_lifecycle_pb.EntryIdentity {
if entry == nil {
return nil
}
id := &s3_lifecycle_pb.EntryIdentity{}
if entry.Attributes != nil {
// FuseAttributes splits the timestamp across Mtime (seconds) and
// MtimeNs (nanosecond component); EntryIdentity.MtimeNs is the
// combined nanoseconds-since-epoch value.
id.MtimeNs = entry.Attributes.Mtime*int64(1e9) + int64(entry.Attributes.MtimeNs)
id.Size = int64(entry.Attributes.FileSize)
}
if len(entry.GetChunks()) > 0 {
id.HeadFid = entry.GetChunks()[0].GetFileIdString()
}
id.ExtendedHash = s3lifecycle.HashExtended(entry.Extended)
return id
}
func identityMatches(live, want *s3_lifecycle_pb.EntryIdentity) bool {
if want == nil {
// No CAS witness (early bootstrap); skip.
return true
}
if live == nil {
return false
}
if live.MtimeNs != want.MtimeNs || live.Size != want.Size {
return false
}
if live.HeadFid != want.HeadFid {
return false
}
return bytes.Equal(live.ExtendedHash, want.ExtendedHash)
}
func done() *s3_lifecycle_pb.LifecycleDeleteResponse {
return &s3_lifecycle_pb.LifecycleDeleteResponse{Outcome: s3_lifecycle_pb.LifecycleDeleteOutcome_DONE}
}
func noopResolved(reason string) *s3_lifecycle_pb.LifecycleDeleteResponse {
return &s3_lifecycle_pb.LifecycleDeleteResponse{Outcome: s3_lifecycle_pb.LifecycleDeleteOutcome_NOOP_RESOLVED, Reason: reason}
}
func blocked(reason string) *s3_lifecycle_pb.LifecycleDeleteResponse {
return &s3_lifecycle_pb.LifecycleDeleteResponse{Outcome: s3_lifecycle_pb.LifecycleDeleteOutcome_BLOCKED, Reason: reason}
}
func retryLater(reason string) *s3_lifecycle_pb.LifecycleDeleteResponse {
return &s3_lifecycle_pb.LifecycleDeleteResponse{Outcome: s3_lifecycle_pb.LifecycleDeleteOutcome_RETRY_LATER, Reason: reason}
}