Files
seaweedfs/weed/s3api/s3api_internal_lifecycle.go
T
Chris LuandClaude Opus 5.5 63806b7073 s3 lifecycle: bound each LifecycleDelete attempt with its own deadline (#12646)
* s3 lifecycle: bound each LifecycleDelete attempt with its own deadline

Replay and walker deletes used the job context with no request
deadline, so one stalled unary RPC could hold a shard for the whole
job. Each attempt now runs under DeleteTimeout (default 1m), set after
the limiter wait. A timeout while the job is still live is retried like
a transport error and then halts the shard or walk without advancing;
parent cancellation still stops both the attempt and the backoff.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: admin setting for the delete RPC timeout

Adds delete_timeout_seconds to the S3 Lifecycle admin form and passes
it to both the replay drain and the walker. 0 keeps the 60s default.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: describe replay and walker timeouts separately in the help text

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: fail the pass when a replay delete exhausts its attempts

A transport error or attempt timeout that outlived the retries halted
the shard with a nil error, so the job reported success. processMatches
now returns the error (the cursor still holds at the event), and
runShard treats a drain error as a clean stop only when its own context
ended, since an attempt timeout also wraps DeadlineExceeded.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: test that a replayed LifecycleDelete resolves after a lost reply

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: skip the expiration write once the attempt deadline passed

A worker retries a LifecycleDelete whose attempt timed out, while the
first attempt can still be running on the S3 server. An attempt that
stalled before its marker write would then stack a second delete marker
after the retry wrote one. The gRPC deadline reaches the handler context,
so check it right before the expiration writes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: recheck the attempt deadline after the suspended null delete

The null-version delete can block on the owner filer past the attempt
deadline; the worker may have retried and written a marker by then.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: finish the suspended marker once the null version is deleted

Skipping the marker after a successful null delete left an older version
as the latest, and the retry then resolved as a no-op without hiding it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3: let a routed delete marker commit only while its latest guard holds

createDeleteMarkerIf passes an optional condition, evaluated on the
.versions directory entry under the object's lock on the owner filer,
and reports a refused commit as errLatestVersionChanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: write the expiration marker as a compare-and-set on the latest pointer

A timed-out attempt can still be running on the S3 server while the
worker's retry writes the marker. The marker commit is now conditioned
on the .versions pointer as read just before it and skipped when the
latest is already a delete marker, so whichever attempt commits second
writes nothing.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: resolve the marker as stale when another version became latest

The marker's compare-and-set only covered changes after its own pointer
read, so a version promoted after the expiration checks passed could be
hidden by the marker. The enabled path now passes the validated version
and a conflict retry that finds a different named latest resolves the
request as VERSION_NOT_LATEST.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: resolve a null expiration as stale once a version takes the pointer

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: read the marker's latest version the way getLatestObjectVersion does

A pointer names a version only with its file name; an id alone falls back
to the null object, and the null-is-latest signal makes the null version
current. An expiration that validated a named version no longer hides a
null written by a suspended PUT, and a current null under an id-only pointer
now expires instead of resolving as stale.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: hold the object write lock for an unrouted expiration marker

Off the ring createDeleteMarkerIf cannot carry the pointer guard, so a PUT
committing between the pointer read and the marker write was hidden. The
read and the write now run under the object write lock that unrouted
writers take.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: share transient-error retries between walker and replay deletes (#12649)

* s3 lifecycle: retry only transient LifecycleDelete failures

dispatchWithRetry retried every RPC error except parent cancellation, so
an authentication, permission or argument error spent the whole retry
budget before surfacing. Retry only attempt timeouts and the failures
util.IsTransientError recognizes (Unavailable, ResourceExhausted, network
resets); everything else returns on the first attempt. A nil response
is now an error instead of a nil dereference.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: take a delete-rate token for every LifecycleDelete attempt

processMatches waited on the shared limiter once per match, so the
retries behind it ran outside the cluster-wide deletes/sec budget. The
retry loop now waits on the limiter before each attempt.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: give walker deletes the replay retry policy

WalkerDispatcher sent each LifecycleDelete once, so one dropped
connection aborted the bucket's daily walk while the same failure on the
replay path was retried. The walker now goes through deleteWithRetry and
shares its transient-error retries, per-attempt limiter wait and nil
response handling.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: retry a DeadlineExceeded the server sends back

A server or proxy can end the RPC with DeadlineExceeded while the
attempt's own deadline is still running. util.IsTransientError rejects
that status, so the delete halted with retries left. It is a timed-out
attempt like the client-side one, so retry it the same way.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* s3 lifecycle: say walker deletes get 3 attempts in the timeout help

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-15 13:09:22 -08:00

566 lines
24 KiB
Go

package s3api
import (
"bytes"
"context"
"encoding/hex"
"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/s3err"
"github.com/seaweedfs/seaweedfs/weed/s3api/s3lifecycle"
stats_collect "github.com/seaweedfs/seaweedfs/weed/stats"
)
// 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 err := s3a.checkAdminAuth(ctx); err != nil {
return nil, err
}
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)
}
// LifecycleReady authorizes the caller as LifecycleDelete does and touches
// no bucket or object metadata.
func (s3a *S3ApiServer) LifecycleReady(ctx context.Context, _ *s3_lifecycle_pb.LifecycleReadyRequest) (*s3_lifecycle_pb.LifecycleReadyResponse, error) {
if err := s3a.checkAdminAuth(ctx); err != nil {
return nil, err
}
return &s3_lifecycle_pb.LifecycleReadyResponse{}, nil
}
func (s3a *S3ApiServer) lifecycleDispatch(ctx context.Context, req *s3_lifecycle_pb.LifecycleDeleteRequest, entry *filer_pb.Entry) (*s3_lifecycle_pb.LifecycleDeleteResponse, error) {
metadataOnly := entryUsesMetadataOnlyDelete(entry)
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
}
// An unaddressed read already resolves the latest when enabled.
if state == s3_constants.VersioningSuspended || (state == s3_constants.VersioningEnabled && req.VersionId != "") {
if resp := s3a.lifecycleRequireLatest(req, entry); resp != nil {
return resp, nil
}
}
// Past the attempt deadline the worker may already be resending
// this request, so a late marker write would stack a second one.
if err := ctx.Err(); err != nil {
return nil, err
}
switch state {
case s3_constants.VersioningEnabled:
written, resp := s3a.lifecycleDeleteMarker(req.Bucket, req.ObjectPath, lifecycleAddressedVersion(req, entry))
if resp != nil {
return resp, nil
}
if !written {
return noopResolved("LATEST_IS_DELETE_MARKER"), nil
}
return done(), nil
case s3_constants.VersioningSuspended:
// Best-effort null delete; NotFound is benign. Route it when an
// owner resolves — on the default filer an absent entry deletes as
// success, and a null version that only reached the owner would
// linger. Lifecycle never bypasses retention.
nullDeleted := false
if owner := s3a.routableWriteOwner(req.Bucket, req.ObjectPath); owner != "" {
worm, lockErr := s3a.isObjectLockEnabled(req.Bucket)
if lockErr == nil {
code, settled := s3a.routedDeleteNullVersion(owner, req.Bucket, req.ObjectPath, worm, false, metadataOnly, nil)
if settled {
if code == s3err.ErrAccessDenied {
return retryLater("ACCESS_DENIED: deleteNullVersion: object lock"), nil
}
if code != s3err.ErrNone {
return retryLater("TRANSPORT_ERROR: routedDeleteNullVersion: " + s3err.GetAPIError(code).Code), nil
}
nullDeleted = true
}
}
}
if !nullDeleted {
if err := s3a.deleteSpecificObjectVersion(ctx, req.Bucket, req.ObjectPath, "null", metadataOnly); err != nil {
if !errors.Is(err, filer_pb.ErrNotFound) && !errors.Is(err, ErrVersionNotFound) {
return retryLater("TRANSPORT_ERROR: deleteNullVersion: " + err.Error()), nil
}
}
}
// The null delete is done, so the marker must follow even past
// the deadline: without it a retry finds an older version latest.
if _, resp := s3a.lifecycleDeleteMarker(req.Bucket, req.ObjectPath, ""); resp != nil {
return resp, nil
}
return done(), nil
default:
err := s3a.WithFilerClient(false, func(c filer_pb.SeaweedFilerClient) error {
return s3a.deleteUnversionedObjectWithClient(ctx, c, req.Bucket, req.ObjectPath, metadataOnly)
})
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
}
recordMetadataOnlyIf(metadataOnly, req)
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
}
}
// Route the version delete when an owner resolves: its recompute
// repoints .versions while the file is removed, all on the filer the
// routed reads consult. A routed failure retries rather than falling
// back, where an absent entry would delete as success while the owner
// still holds the version. Lifecycle never bypasses retention.
versionDeleted := false
if owner := s3a.routableWriteOwner(req.Bucket, req.ObjectPath); owner != "" {
worm, lockErr := s3a.isObjectLockEnabled(req.Bucket)
if lockErr == nil {
// The "null" version is the object entry, not a .versions file.
if req.VersionId == "null" {
code, settled := s3a.routedDeleteNullVersion(owner, req.Bucket, req.ObjectPath, worm, false, metadataOnly, nil)
if settled {
if code == s3err.ErrAccessDenied {
return retryLater("ACCESS_DENIED: deleteSpecificVersion: object lock"), nil
}
if code != s3err.ErrNone {
return retryLater("TRANSPORT_ERROR: routedDeleteNullVersion: " + s3err.GetAPIError(code).Code), nil
}
versionDeleted = true
}
} else {
switch code := s3a.routedDeleteSpecificVersion(owner, req.Bucket, req.ObjectPath, req.VersionId, worm && !isDeleteMarkerEntry(entry), false, metadataOnly, nil); code {
case s3err.ErrNone:
versionDeleted = true
case s3err.ErrAccessDenied:
return retryLater("ACCESS_DENIED: deleteSpecificVersion: object lock"), nil
default:
return retryLater("TRANSPORT_ERROR: routedDeleteSpecificVersion: " + s3err.GetAPIError(code).Code), nil
}
}
}
}
if !versionDeleted {
if err := s3a.deleteSpecificObjectVersion(ctx, req.Bucket, req.ObjectPath, req.VersionId, metadataOnly); 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
}
}
recordMetadataOnlyIf(metadataOnly, req)
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
}
}
// lifecycleDeleteMarker writes an expiration's delete marker as a
// compare-and-set on the .versions pointer, skipping it when the latest is
// already a delete marker. An attempt the worker gave up on may still be
// running while its retry writes, so whichever commits second sees the marker.
// A non-empty expected is the version that passed the expiration checks; a
// different latest resolves the request as stale instead of hiding it.
func (s3a *S3ApiServer) lifecycleDeleteMarker(bucket, object, expected string) (written bool, resp *s3_lifecycle_pb.LifecycleDeleteResponse) {
normalized := s3_constants.NormalizeObjectKey(object)
if s3a.objectWriteOwner(bucket, normalized) != "" {
return s3a.lifecycleDeleteMarkerIf(bucket, object, normalized, expected)
}
// Off the ring the marker commits without the guard, so the pointer read
// and the write serialize with other writers on the object write lock.
code := s3a.withObjectWriteLock(bucket, normalized, nil, func() s3err.ErrorCode {
written, resp = s3a.lifecycleDeleteMarkerIf(bucket, object, normalized, expected)
return s3err.ErrNone
})
if code != s3err.ErrNone {
return false, retryLater("TRANSPORT_ERROR: object write lock: " + s3err.GetAPIError(code).Code)
}
return written, resp
}
func (s3a *S3ApiServer) lifecycleDeleteMarkerIf(bucket, object, normalized, expected string) (bool, *s3_lifecycle_pb.LifecycleDeleteResponse) {
for range 3 {
pointer, err := s3a.getObjectScopedEntryRoutedByKey(bucket, normalized, s3a.bucketDir(bucket), normalized+s3_constants.VersionsFolder)
if err != nil {
if !isObjectEntryMiss(err) {
return false, retryLater("TRANSPORT_ERROR: latest-pointer lookup: " + err.Error())
}
pointer = nil
}
if pointer != nil && string(pointer.Extended[s3_constants.ExtLatestVersionIsDeleteMarker]) == "true" && !nullVersionIsLatest(pointer) {
return false, nil
}
if expected != "" && pointer != nil {
if latest := pointerLatestVersion(pointer); latest != "" && latest != expected {
return false, noopResolved("VERSION_NOT_LATEST")
}
}
guard := &filer_pb.WriteCondition{Clauses: []*filer_pb.WriteCondition_Clause{{
Kind: filer_pb.WriteCondition_IF_ENTRY_EQUAL,
ExpectedEntry: pointer,
}}}
if _, err := s3a.createDeleteMarkerIf(bucket, object, guard); err != nil {
if errors.Is(err, errLatestVersionChanged) {
continue
}
return false, retryLater("TRANSPORT_ERROR: createDeleteMarker: " + err.Error())
}
return true, nil
}
return false, retryLater("LATEST_CHANGED: createDeleteMarker")
}
// pointerLatestVersion reads the latest version off a .versions pointer the
// way getLatestObjectVersion does: a named pointer with its file outranks the
// null object, the null-is-latest signal names the null version, and anything
// else leaves the latest to the caller's earlier read.
func pointerLatestVersion(pointer *filer_pb.Entry) string {
id := string(pointer.Extended[s3_constants.ExtLatestVersionIdKey])
if id != "" && len(pointer.Extended[s3_constants.ExtLatestVersionFileNameKey]) > 0 {
return id
}
if nullVersionIsLatest(pointer) {
return "null"
}
return ""
}
// lifecycleRequireLatest refuses a current-version expiration whose version
// is no longer the key's latest. A version file never changes, so the identity
// check cannot see a newer write.
func (s3a *S3ApiServer) lifecycleRequireLatest(req *s3_lifecycle_pb.LifecycleDeleteRequest, entry *filer_pb.Entry) *s3_lifecycle_pb.LifecycleDeleteResponse {
latest, err := s3a.getLatestObjectVersion(req.Bucket, req.ObjectPath)
if err != nil {
if isObjectEntryMiss(err) {
return noopResolved("NOT_FOUND")
}
return retryLater("TRANSPORT_ERROR: latest-version lookup: " + err.Error())
}
if versionIdOrNull(latest) != lifecycleAddressedVersion(req, entry) {
return noopResolved("VERSION_NOT_LATEST")
}
return nil
}
func versionIdOrNull(e *filer_pb.Entry) string {
if id := string(e.Extended[s3_constants.ExtVersionIdKey]); id != "" {
return id
}
return "null"
}
func lifecycleAddressedVersion(req *s3_lifecycle_pb.LifecycleDeleteRequest, entry *filer_pb.Entry) string {
if req.VersionId != "" {
return req.VersionId
}
return versionIdOrNull(entry)
}
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.
uploadEntry, err := s3a.getEntry(uploadsFolder, uploadID)
if err != nil {
if errors.Is(err, filer_pb.ErrNotFound) {
return noopResolved("NOT_FOUND"), nil
}
return retryLater("TRANSPORT_ERROR: getEntry: " + err.Error()), nil
}
if uploadEntry == nil {
return noopResolved("NOT_FOUND"), nil
}
object := string(uploadEntry.Extended[s3_constants.ExtMultipartObjectKey])
code := s3a.withObjectWriteLock(req.Bucket, object, nil, func() s3err.ErrorCode {
return s3a.removeUploadDir(req.Bucket, uploadID, object)
})
if code == s3err.ErrNone {
return done(), nil
}
return retryLater("TRANSPORT_ERROR: removeUploadDir: " + s3err.GetAPIError(code).Code), 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)
}
// entryUsesMetadataOnlyDelete reports whether the lifecycle delete path
// can skip per-chunk DeleteFile RPCs and rely on the volume's TTL to
// reclaim chunks. Per-write TTL stamping (PR 9377) sets Attributes.TtlSec
// on every entry whose lifecycle rule fits within volume TTL — observing
// a non-zero TtlSec on the live entry is the authoritative signal.
// Defensive nil-checks because the caller may be racing a concurrent
// rewrite that nil-ed Attributes briefly during meta-log replay.
func entryUsesMetadataOnlyDelete(entry *filer_pb.Entry) bool {
return entry != nil && entry.Attributes != nil && entry.Attributes.TtlSec > 0
}
// recordMetadataOnlyIf bumps the metadata-only counter when on=true.
// Skipped when off so callers don't need a guard at every call site.
// rule_hash is hex-encoded so operators can group by rule when
// debugging; nil rule_hash collapses to the empty string.
func recordMetadataOnlyIf(on bool, req *s3_lifecycle_pb.LifecycleDeleteRequest) {
if !on || req == nil {
return
}
stats_collect.S3LifecycleMetadataOnlyCounter.WithLabelValues(req.Bucket, hex.EncodeToString(req.RuleHash)).Inc()
}
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}
}