fix(ilm): report refused purges of all versions, and account their keys

An expiry of all versions of a key purges exactly that key, and the
purge can refuse it or stop early. Those outcomes were hard to see, and
the scanner reported a key it had queued as if it were already gone.

- Refusals and failures of a purge are counted in
  expiry_incomplete_total with their own reasons: tier_unsupported (a
  version is in a remote tier), budget (stopped after progress; the next
  expiry continues), changed (the latest version a batch matched
  changed) and history_unavailable. They are traced with the reason and
  send no removal event. A partial purge keeps its audit record, now
  with the reason. A tier refusal is also logged, once per key and hour,
  in words.
- A drive that cannot be read or decoded during the purge's own drive
  check now returns errDeleteAllHistoryUnavailable, which names the
  cause and stays a write quorum error (503) for API callers.
- The scanner does not queue an expiry of all versions whose latest
  version is hot when the pool holds a version in a remote tier: the
  purge would refuse it each cycle. It counts each such observation as
  tier_unsupported_skipped, once per drive scanner and cycle. A tiered
  version only another pool holds is unknown here, so the purge still
  decides; a transitioned latest version keeps the transitioned expiry.
  The tier predicate is shared with the purge's own check, for FileInfo
  and ObjectInfo, and excludes free versions.
- The scanner accounts every version of a key under an expiry of all
  versions at its size, queued or not, instead of as already removed.
  Usage, and so hard quota, drops after a fresh scan of the subtree once
  the purge succeeded. The alerts on many or large versions count the
  versions of a key that is not queued.
- A lifecycle expiry never deletes a prefix: DeleteObject refuses
  DeletePrefix with Expiration.Expire as InvalidArgument. No caller
  sends it since expiries of all versions became purges; a force delete
  of a prefix is unchanged.

Not done from the plan: queuing an MRF heal per version when a drive's
xl.meta fails to decode. Object heal does not rewrite metadata that
fails its CRC check (it reports the drive as unknown and skips it), so
those heals would not repair the drive. Such a key stays refused until
its metadata on that drive is repaired.

Tests: TestPurgeAllVersionsRefusalReasons,
TestPurgeAllVersionsRefusesCorruptDrive,
TestScannerSkipsDeleteAllForTieredKey and
TestPoolsDeleteObjectRejectsExpirePrefix are new. Tests that expected
the scanner to queue a refused purge, the predictive accounting, or
the old prefix options are converted.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Feng Ruohang <rh@vonng.com>

cmd/data-scanner.go

@@ -24,10 +24,12 @@ import (
 	"errors"
 	"fmt"
 	"io/fs"
+	"maps"
 	"math"
 	"math/rand"
 	"os"
 	"path"
+	"slices"
 	"strconv"
 	"strings"
 	"sync"
@@ -1118,8 +1120,31 @@ eventLoop:
 		size := actualSz
 		switch event.Action {
 		case lifecycle.DeleteAllVersionsAction, lifecycle.DelMarkerDeleteAllVersionsAction:
-			remainingVersions = 0
-			applyExpiryRule(event, lcEventSrc_Scanner, oi)
+			// An expiry of a hot latest version purges the key, which refuses
+			// a key with a version in a remote tier: do not queue it when this
+			// pool holds one. A tiered version only another pool holds is
+			// not known here; the purge decides on it. A transitioned latest
+			// version takes the transitioned expiry, as before.
+			refused := oi.TransitionedObject.Status == "" && slices.ContainsFunc(objInfos, ObjectInfo.deleteAllTierRefuses)
+			if refused {
+				globalILMIncomplete.add(ilmIncompleteTierSkipped)
+			} else {
+				remainingVersions = 0
+				applyExpiryRule(event, lcEventSrc_Scanner, oi)
+			}
+			// The versions hold their space until the purge removes them,
+			// which can refuse the key or stop at its budget: account each
+			// at its size, with no other action. The alerts on many or
+			// large versions count those of a key that is not queued.
+			for _, v := range objInfos[idx:] {
+				sz, _ := v.GetActualSize()
+				if fn != nil {
+					fn(v, sz, sz, sizeS)
+				}
+				if refused {
+					cumulativeSize += sz
+				}
+			}
 			break eventLoop
 
 		case lifecycle.DeleteAction, lifecycle.DeleteRestoredAction, lifecycle.DeleteRestoredVersionAction:
@@ -1320,16 +1345,24 @@ const (
 // before it deleted anything, reports what a rejected expiry does. One that
 // failed or stopped after it began to delete is audited as partial, with the
 // versions it removed, but sends no removal event: the key is still there.
+//
+// Each is counted and traced with its reason, as reportILMExpiryNotDone
+// reports other expiries. A refusal because a version is in a remote tier is
+// also logged, since it lasts until such versions can be purged.
 func deleteAllNotCompleted(ctx context.Context, obj ObjectInfo, lcEvent lifecycle.Event, src lcEventSrc, purge deleteAllResult, err error, traceFn func(event string, metadata map[string]string, err error)) bool {
+	logID := "non-transition-expiry" + obj.Name
 	switch {
 	case err == nil: // there was nothing to remove
 		traceFn(ILMExpiry, nil, nil)
 		return false
 	case purge.Started:
+		reason, _ := ilmExpiryNotDone(err)
+		globalILMIncomplete.add(reason)
 		tags := newLifecycleAuditEvent(src, lcEvent).Tags()
 		tags["version-id"] = purge.Latest.VersionID
 		tags[ilmVersionsRemoved] = strconv.Itoa(purge.Removed)
 		tags[ilmPartial] = "true"
+		tags["ilm-incomplete-reason"] = reason
 		auditLogInternal(ctx, AuditLogOptions{
 			Event:     ILMExpiry,
 			APIName:   "ILMExpiry",
@@ -1339,22 +1372,22 @@ func deleteAllNotCompleted(ctx context.Context, obj ObjectInfo, lcEvent lifecycl
 			Error:     err.Error(),
 			Tags:      tags,
 		})
-		traceFn(ILMExpiry, tags, err)
+		trace := maps.Clone(tags)
+		trace["result"], trace["reason"] = "partial", reason
+		traceFn(ILMExpiry, trace, err)
 		if errors.Is(err, errDeleteAllBudget) {
 			return false // a long history; the next expiry continues
 		}
+		// Assume it is still there.
+		ilmLogOnceIf(ctx, fmt.Errorf("ILM expiry of all versions of %s/%s stopped after it removed %d: %w", obj.Bucket, obj.Name, purge.Removed, err), logID)
+		return false
 	case errors.Is(err, errDeleteAllTiered):
 		// Kept until tiered versions can be purged; see deleteAllVersions.
-		traceFn(ILMExpiry, nil, err)
-		return false
-	case errors.Is(err, errLifecycleNotApplicable) || isErrObjectNotFound(err) || isErrVersionNotFound(err):
-		traceFn(ILMExpiry, nil, nil)
-		return false
-	default:
-		traceFn(ILMExpiry, nil, err)
+		ilmLogOnceIf(ctx, fmt.Errorf("ILM expiry of all versions of %s/%s refused: a version is in a remote tier, and a purge of all versions does not remove tiered versions yet", obj.Bucket, obj.Name), logID)
+	}
+	if !reportILMExpiryNotDone(ctx, obj, lcEvent, src, err, logID, traceFn) {
+		traceFn(ILMExpiry, nil, nil) // gone already
 	}
-	// Assume it is still there.
-	ilmLogOnceIf(ctx, fmt.Errorf("DeleteObject(%s, %s): %w", obj.Bucket, obj.Name, err), "non-transition-expiry"+obj.Name)
 	return false
 }