fix: avoid waiting on rebalance metadata (#20392)

rebalance metadata is good to have only,
if it cannot be loaded when starting MinIO
for some reason we can possibly ignore it
and move on and let user start rebalance
again if needed.

cmd/erasure-server-pool-decom.go

@@ -1,4 +1,4 @@
-// Copyright (c) 2015-2023 MinIO, Inc.
+// Copyright (c) 2015-2024 MinIO, Inc.
 //
 // This file is part of MinIO Object Storage stack
 //
@@ -490,14 +490,11 @@ const (
 // in 'pool.bin', this is eventually used for decommissioning the pool.
 func (z *erasureServerPools) Init(ctx context.Context) error {
 	// Load rebalance metadata if present
-	err := z.loadRebalanceMeta(ctx)
-	if err != nil {
-		return fmt.Errorf("failed to load rebalance data: %w", err)
+	if err := z.loadRebalanceMeta(ctx); err == nil {
+		// Start rebalance routine if we can reload rebalance metadata.
+		z.StartRebalance()
 	}
 
-	// Start rebalance routine
-	z.StartRebalance()
-
 	meta := poolMeta{}
 	if err := meta.load(ctx, z.serverPools[0], z.serverPools); err != nil {
 		return err

cmd/erasure-server-pool-rebalance.go

@@ -110,19 +110,14 @@ var errRebalanceNotStarted = errors.New("rebalance not started")
 
 func (z *erasureServerPools) loadRebalanceMeta(ctx context.Context) error {
 	r := &rebalanceMeta{}
-	err := r.load(ctx, z.serverPools[0])
-	if err != nil {
-		if errors.Is(err, errConfigNotFound) {
-			return nil
-		}
-		return err
+	if err := r.load(ctx, z.serverPools[0]); err == nil {
+		z.rebalMu.Lock()
+		z.rebalMeta = r
+		z.updateRebalanceStats(ctx)
+		z.rebalMu.Unlock()
+	} else if !errors.Is(err, errConfigNotFound) {
+		rebalanceLogIf(ctx, fmt.Errorf("failed to load rebalance metadata, continue to restart rebalance as needed: %w", err))
 	}
-
-	z.rebalMu.Lock()
-	z.rebalMeta = r
-	z.updateRebalanceStats(ctx)
-	z.rebalMu.Unlock()
-
 	return nil
 }