Report job success immediately after main graph completes, before post steps

- Report success to users as soon as main graph completes successfully
- Post steps (promotion/cleanup) now run as best-effort and don't affect job result
- Add metrics tracking: main_graph_duration_seconds, post_steps_duration_seconds, time_saved_seconds
- Metrics are extractable from ci-operator-metrics.json in test_platform_insights events

Author: deepsm007

cmd/ci-operator/main.go

@@ -1057,26 +1057,55 @@ func (o *options) Run() []error {
 			return wrapped
 		}
 
-		// Run each of the promotion steps concurrently
+		// Main graph completed successfully - report success immediately before post steps
+		mainGraphCompletedAt := time.Now()
+		mainGraphDuration := mainGraphCompletedAt.Sub(start)
+		eventRecorder.Event(runtimeObject, coreapi.EventTypeNormal, "CiJobSucceeded", eventJobDescription(o.jobSpec, o.namespace))
+
+		// Report success to users immediately (post steps are best-effort cleanup)
+		reporter, loadErr := o.resultsOptions.Reporter(o.jobSpec, o.consoleHost)
+		if loadErr != nil {
+			logrus.WithError(loadErr).Warn("Could not load result reporting options, skipping early success report.")
+		} else {
+			reporter.Report(nil)
+		}
+
+		// Run each of the promotion steps concurrently (best-effort cleanup)
+		postStepsStart := time.Now()
 		lenOfPromotionSteps := len(promotionSteps)
 		detailsChan := make(chan api.CIOperatorStepDetails, lenOfPromotionSteps)
 		errChan := make(chan error, lenOfPromotionSteps)
 		for _, step := range promotionSteps {
 			go runPromotionStep(ctx, step, detailsChan, errChan, o.metricsAgent)
 		}
+		postStepsFailed := false
 		for i := 0; i < lenOfPromotionSteps; i++ {
 			select {
 			case details := <-detailsChan:
 				graph.MergeFrom(details)
 			case err := <-errChan:
 				errorDesc := fmt.Sprintf("post step failed while %s. with error: %v", eventJobDescription(o.jobSpec, o.namespace), err)
 				eventRecorder.Event(runtimeObject, coreapi.EventTypeWarning, "PostStepFailed", errorDesc)
-				return []error{results.ForReason("executing_post").WithError(err).Unwrap()} // If any of the promotion steps fail, it is considered a failure
+				logrus.WithError(err).Warn("Post step failed, but job success was already reported. Continuing with cleanup.")
+				postStepsFailed = true
+				// Post step failures don't affect job success (already reported), but we still record them
 			}
 		}
 
-		eventRecorder.Event(runtimeObject, coreapi.EventTypeNormal, "CiJobSucceeded", eventJobDescription(o.jobSpec, o.namespace))
-		o.metricsAgent.Record(metrics.NewInsightsEvent(metrics.InsightExecutionCompleted, metrics.Context{"duration_seconds": time.Since(start).Seconds(), "success": true}))
+		// Record final metrics including post steps duration
+		postStepsDuration := time.Since(postStepsStart)
+		totalDuration := time.Since(start)
+		metricsContext := metrics.Context{
+			"duration_seconds":            totalDuration.Seconds(),
+			"main_graph_duration_seconds": mainGraphDuration.Seconds(),
+			"post_steps_duration_seconds": postStepsDuration.Seconds(),
+			"time_saved_seconds":          postStepsDuration.Seconds(),
+			"success":                     true,
+		}
+		if postStepsFailed {
+			metricsContext["post_steps_failed"] = true
+		}
+		o.metricsAgent.Record(metrics.NewInsightsEvent(metrics.InsightExecutionCompleted, metricsContext))
 
 		return nil
 	})