heal: Reset healing params when a retry is decided (#20285)

Currently, retry healing of a new drive healing does not reset
HealedBuckets means that the next healing retry will skip those
buckets. The commit will fix this behavior.

Also, the skipped objects counter will include objects uploaded
that are uploaded after the healing is started.

cmd/background-newdisks-heal-ops.go

@@ -148,6 +148,26 @@ func initHealingTracker(disk StorageAPI, healID string) *healingTracker {
 	return h
 }
 
+func (h *healingTracker) resetHealing() {
+	h.mu.Lock()
+	defer h.mu.Unlock()
+
+	h.ItemsHealed = 0
+	h.ItemsFailed = 0
+	h.BytesDone = 0
+	h.BytesFailed = 0
+	h.ResumeItemsHealed = 0
+	h.ResumeItemsFailed = 0
+	h.ResumeBytesDone = 0
+	h.ResumeBytesFailed = 0
+	h.ItemsSkipped = 0
+	h.BytesSkipped = 0
+
+	h.HealedBuckets = nil
+	h.Object = ""
+	h.Bucket = ""
+}
+
 func (h *healingTracker) getLastUpdate() time.Time {
 	h.mu.RLock()
 	defer h.mu.RUnlock()
@@ -349,6 +369,7 @@ func (h *healingTracker) toHealingDisk() madmin.HealingDisk {
 		Object:            h.Object,
 		QueuedBuckets:     h.QueuedBuckets,
 		HealedBuckets:     h.HealedBuckets,
+		RetryAttempts:     h.RetryAttempts,
 
 		ObjectsHealed: h.ItemsHealed, // Deprecated July 2021
 		ObjectsFailed: h.ItemsFailed, // Deprecated July 2021
@@ -482,16 +503,19 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 	// if objects have failed healing, we attempt a retry to heal the drive upto 3 times before giving up.
 	if tracker.ItemsFailed > 0 && tracker.RetryAttempts < 4 {
 		tracker.RetryAttempts++
-		bugLogIf(ctx, tracker.update(ctx))
 
 		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retrying %s time (healed: %d, skipped: %d, failed: %d).", disk,
 			humanize.Ordinal(int(tracker.RetryAttempts)), tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+
+		tracker.resetHealing()
+		bugLogIf(ctx, tracker.update(ctx))
+
 		return errRetryHealing
 	}
 
 	if tracker.ItemsFailed > 0 {
 		healingLogEvent(ctx, "Healing of drive '%s' is incomplete, retried %d times (healed: %d, skipped: %d, failed: %d).", disk,
-			tracker.RetryAttempts-1, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+			tracker.RetryAttempts, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
 	} else {
 		if tracker.RetryAttempts > 0 {
 			healingLogEvent(ctx, "Healing of drive '%s' is complete, retried %d times (healed: %d, skipped: %d).", disk,

cmd/global-heal.go

@@ -167,6 +167,19 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 		return errServerNotInitialized
 	}
 
+	started := tracker.Started
+	if started.IsZero() || started.Equal(timeSentinel) {
+		healingLogIf(ctx, fmt.Errorf("unexpected tracker healing start time found: %v", started))
+		started = time.Time{}
+	}
+
+	// Final tracer update before quitting
+	defer func() {
+		tracker.setObject("")
+		tracker.setBucket("")
+		healingLogIf(ctx, tracker.update(ctx))
+	}()
+
 	for _, bucket := range healBuckets {
 		if err := bgSeq.healBucket(objAPI, bucket, true); err != nil {
 			// Log bucket healing error if any, we shall retry again.
@@ -435,13 +448,10 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 
 			var versionNotFound int
 			for _, version := range fivs.Versions {
-				// Ignore a version with a modtime newer than healing start time.
+				// Ignore healing a version if:
-				if version.ModTime.After(tracker.Started) {
+				// - It is uploaded after the drive healing is started
-					continue
+				// - An object that is already expired by ILM rule.
-				}
+				if !started.IsZero() && version.ModTime.After(started) || filterLifecycle(bucket, version.Name, version) {
-
-				// Apply lifecycle rules on the objects that are expired.
-				if filterLifecycle(bucket, version.Name, version) {
 					versionNotFound++
 					if !send(healEntrySkipped(uint64(version.Size))) {
 						return
@@ -556,10 +566,6 @@ func (er *erasureObjects) healErasureSet(ctx context.Context, buckets []string,
 			healingLogIf(ctx, tracker.update(ctx))
 		}
 	}
-
-	tracker.setObject("")
-	tracker.setBucket("")
-
 	if retErr != nil {
 		return retErr
 	}

go.mod

@@ -51,7 +51,7 @@ require (
 	github.com/minio/highwayhash v1.0.3
 	github.com/minio/kms-go/kes v0.3.0
 	github.com/minio/kms-go/kms v0.4.0
-	github.com/minio/madmin-go/v3 v3.0.63
+	github.com/minio/madmin-go/v3 v3.0.64-0.20240822003756-fe52a32e526d
 	github.com/minio/minio-go/v7 v7.0.75
 	github.com/minio/mux v1.9.0
 	github.com/minio/pkg/v3 v3.0.11

go.sum

@@ -426,8 +426,8 @@ github.com/minio/kms-go/kes v0.3.0 h1:SU8VGVM/Hk9w1OiSby3OatkcojooUqIdDHl6dtM6Nk
 github.com/minio/kms-go/kes v0.3.0/go.mod h1:w6DeVT878qEOU3nUrYVy1WOT5H1Ig9hbDIh698NYJKY=
 github.com/minio/kms-go/kms v0.4.0 h1:cLPZceEp+05xHotVBaeFJrgL7JcXM4lBy6PU0idkE7I=
 github.com/minio/kms-go/kms v0.4.0/go.mod h1:q12CehiIy2qgBnDKq6Q7wmPi2PHSyRVug5DKp0HAVeE=
-github.com/minio/madmin-go/v3 v3.0.63 h1:ERJRxEI/FFRh8MDi4Z+3DKe4sONkQ0g+OkNzRpk7qxk=
+github.com/minio/madmin-go/v3 v3.0.64-0.20240822003756-fe52a32e526d h1:ma9PAmbEs+TP9BdsbQLO3gUa2nHSzeuQobOCT8BWUpg=
-github.com/minio/madmin-go/v3 v3.0.63/go.mod h1:IFAwr0XMrdsLovxAdCcuq/eoL4nRuMVQQv0iubJANQw=
+github.com/minio/madmin-go/v3 v3.0.64-0.20240822003756-fe52a32e526d/go.mod h1:IFAwr0XMrdsLovxAdCcuq/eoL4nRuMVQQv0iubJANQw=
 github.com/minio/mc v0.0.0-20240815155011-479171e7be9c h1:0tzuJ1nV6oZstqKQ/CwK1dzxNJ/cE38ym4SPi2HsWoY=
 github.com/minio/mc v0.0.0-20240815155011-479171e7be9c/go.mod h1:Cr4x7eiMJfOTWwg40Rk3EaOI7i+DUyOAtqLO7x+heiA=
 github.com/minio/md5-simd v1.1.2 h1:Gdi1DZK69+ZVMoNHRXJyNcxrMA4dSxoYHZSQbirFg34=