diff --git a/CHANGELOG.md b/CHANGELOG.md
index bbf9175ee48..ccbcc962114 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -40,6 +40,7 @@ storage:
 
 ```
 * [CHANGE] Ignore context canceled errors in the queriers [#2440](https://github.com/grafana/tempo/pull/2440) (@joe-elliott)
+* [CHANGE] Start flush queue worker after wal replay and block rediscovery [#2456](https://github.com/grafana/tempo/pull/2456) (@ie-pham)
 
 ## v2.1.1 / 2023-04-28
 * [BUGFIX] Fix issue where Tempo sometimes flips booleans from false->true at storage time. [#2400](https://github.com/grafana/tempo/issues/2400) (@joe-elliott)
diff --git a/modules/ingester/ingester.go b/modules/ingester/ingester.go
index 9af716bfd78..50cd2445299 100644
--- a/modules/ingester/ingester.go
+++ b/modules/ingester/ingester.go
@@ -81,11 +81,6 @@ func New(cfg Config, store storage.Store, limits *overrides.Overrides, reg prome
 
 	i.local = store.WAL().LocalBackend()
 
-	i.flushQueuesDone.Add(cfg.ConcurrentFlushes)
-	for j := 0; j < cfg.ConcurrentFlushes; j++ {
-		go i.flushLoop(j)
-	}
-
 	lc, err := ring.NewLifecycler(cfg.LifecyclerConfig, i, "ingester", cfg.OverrideRingKey, true, log.Logger, prometheus.WrapRegistererWithPrefix("tempo_", reg))
 	if err != nil {
 		return nil, fmt.Errorf("NewLifecycler failed: %w", err)
@@ -114,6 +109,11 @@ func (i *Ingester) starting(ctx context.Context) error {
 		return fmt.Errorf("failed to rediscover local blocks: %w", err)
 	}
 
+	i.flushQueuesDone.Add(i.cfg.ConcurrentFlushes)
+	for j := 0; j < i.cfg.ConcurrentFlushes; j++ {
+		go i.flushLoop(j)
+	}
+
 	// Now that user states have been created, we can start the lifecycler.
 	// Important: we want to keep lifecycler running until we ask it to stop, so we need to give it independent context
 	if err := i.lifecycler.StartAsync(context.Background()); err != nil {