Fix scraped targets metric (#5263)

CHANGELOG.md

@@ -123,6 +123,9 @@ Main (unreleased)
 
 - Fix `loki.source.file` race condition in cleaning up metrics when stopping to tail files. (@thampiotr)
 
+- Fixed the `agent_prometheus_scrape_targets_gauge` incorrectly reporting all discovered targets 
+  instead of targets that belong to current instance when clustering is enabled. (@thampiotr)
+
 v0.36.1 (2023-09-06)
 --------------------
 

component/prometheus/scrape/scrape.go

@@ -217,6 +217,7 @@ func (c *Component) Run(ctx context.Context) error {
 			// 'clustered' targets implementation every time.
 			ct := discovery.NewDistributedTargets(cl, c.cluster, tgs)
 			promTargets := c.componentTargetsToProm(jobName, ct.Get())
+			c.targetsGauge.Set(float64(len(promTargets)))
 
 			select {
 			case targetSetsChan <- promTargets:
@@ -251,7 +252,6 @@ func (c *Component) Update(args component.Arguments) error {
 	default:
 	}
 
-	c.targetsGauge.Set(float64(len(c.args.Targets)))
 	return nil
 }