add healing retries when there are failed heal attempts (#19986)

transient errors for long running tasks are normal, allow for
drive to retry again upto 3 times before giving up on healing
the drive.

cmd/background-newdisks-heal-ops.go

@@ -88,6 +88,8 @@ type healingTracker struct {
 
 	ItemsSkipped uint64
 	BytesSkipped uint64
+
+	RetryAttempts uint64
 	// Add future tracking capabilities
 	// Be sure that they are included in toHealingDisk
 }
@@ -382,6 +384,8 @@ func getLocalDisksToHeal() (disksToHeal Endpoints) {
 
 var newDiskHealingTimeout = newDynamicTimeout(30*time.Second, 10*time.Second)
 
+var errRetryHealing = errors.New("some items failed to heal, we will retry healing this drive again")
+
 func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint) error {
 	poolIdx, setIdx := endpoint.PoolIdx, endpoint.SetIdx
 	disk := getStorageViaEndpoint(endpoint)
@@ -451,8 +455,27 @@ func healFreshDisk(ctx context.Context, z *erasureServerPools, endpoint Endpoint
 		return err
 	}
 
-	healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d, failed: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped, tracker.ItemsFailed)
+	// 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)
+		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)
+	} else {
+		if tracker.RetryAttempts > 0 {
+			healingLogEvent(ctx, "Healing of drive '%s' is complete, retried %d times (healed: %d, skipped: %d).", disk,
+				tracker.RetryAttempts-1, tracker.ItemsHealed, tracker.ItemsSkipped)
+		} else {
+			healingLogEvent(ctx, "Healing of drive '%s' is finished (healed: %d, skipped: %d).", disk, tracker.ItemsHealed, tracker.ItemsSkipped)
+		}
+	}
 	if serverDebugLog {
 		tracker.printTo(os.Stdout)
 		fmt.Printf("\n")
@@ -524,7 +547,7 @@ func monitorLocalDisksAndHeal(ctx context.Context, z *erasureServerPools) {
 					if err := healFreshDisk(ctx, z, disk); err != nil {
 						globalBackgroundHealState.setDiskHealingStatus(disk, false)
 						timedout := OperationTimedOut{}
-						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) {
+						if !errors.Is(err, context.Canceled) && !errors.As(err, &timedout) && !errors.Is(err, errRetryHealing) {
 							printEndpointError(disk, err, false)
 						}
 						return

cmd/background-newdisks-heal-ops_gen.go

@@ -200,6 +200,12 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "BytesSkipped")
 				return
 			}
+		case "RetryAttempts":
+			z.RetryAttempts, err = dc.ReadUint64()
+			if err != nil {
+				err = msgp.WrapError(err, "RetryAttempts")
+				return
+			}
 		default:
 			err = dc.Skip()
 			if err != nil {
@@ -213,9 +219,9 @@ func (z *healingTracker) DecodeMsg(dc *msgp.Reader) (err error) {
 
 // EncodeMsg implements msgp.Encodable
 func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
-	// map header, size 25
+	// map header, size 26
 	// write "ID"
-	err = en.Append(0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	err = en.Append(0xde, 0x0, 0x1a, 0xa2, 0x49, 0x44)
 	if err != nil {
 		return
 	}
@@ -478,15 +484,25 @@ func (z *healingTracker) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "BytesSkipped")
 		return
 	}
+	// write "RetryAttempts"
+	err = en.Append(0xad, 0x52, 0x65, 0x74, 0x72, 0x79, 0x41, 0x74, 0x74, 0x65, 0x6d, 0x70, 0x74, 0x73)
+	if err != nil {
+		return
+	}
+	err = en.WriteUint64(z.RetryAttempts)
+	if err != nil {
+		err = msgp.WrapError(err, "RetryAttempts")
+		return
+	}
 	return
 }
 
 // MarshalMsg implements msgp.Marshaler
 func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
-	// map header, size 25
+	// map header, size 26
 	// string "ID"
-	o = append(o, 0xde, 0x0, 0x19, 0xa2, 0x49, 0x44)
+	o = append(o, 0xde, 0x0, 0x1a, 0xa2, 0x49, 0x44)
 	o = msgp.AppendString(o, z.ID)
 	// string "PoolIndex"
 	o = append(o, 0xa9, 0x50, 0x6f, 0x6f, 0x6c, 0x49, 0x6e, 0x64, 0x65, 0x78)
@@ -566,6 +582,9 @@ func (z *healingTracker) MarshalMsg(b []byte) (o []byte, err error) {
 	// string "BytesSkipped"
 	o = append(o, 0xac, 0x42, 0x79, 0x74, 0x65, 0x73, 0x53, 0x6b, 0x69, 0x70, 0x70, 0x65, 0x64)
 	o = msgp.AppendUint64(o, z.BytesSkipped)
+	// string "RetryAttempts"
+	o = append(o, 0xad, 0x52, 0x65, 0x74, 0x72, 0x79, 0x41, 0x74, 0x74, 0x65, 0x6d, 0x70, 0x74, 0x73)
+	o = msgp.AppendUint64(o, z.RetryAttempts)
 	return
 }
 
@@ -763,6 +782,12 @@ func (z *healingTracker) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "BytesSkipped")
 				return
 			}
+		case "RetryAttempts":
+			z.RetryAttempts, bts, err = msgp.ReadUint64Bytes(bts)
+			if err != nil {
+				err = msgp.WrapError(err, "RetryAttempts")
+				return
+			}
 		default:
 			bts, err = msgp.Skip(bts)
 			if err != nil {
@@ -785,6 +810,6 @@ func (z *healingTracker) Msgsize() (s int) {
 	for za0002 := range z.HealedBuckets {
 		s += msgp.StringPrefixSize + len(z.HealedBuckets[za0002])
 	}
-	s += 7 + msgp.StringPrefixSize + len(z.HealID) + 13 + msgp.Uint64Size + 13 + msgp.Uint64Size
+	s += 7 + msgp.StringPrefixSize + len(z.HealID) + 13 + msgp.Uint64Size + 13 + msgp.Uint64Size + 14 + msgp.Uint64Size
 	return
 }