From 48274ab696d852696c5c28606f5605968b2fbc20 Mon Sep 17 00:00:00 2001
From: vyzaldysanchez <vyzaldysanchez@gmail.com>
Date: Thu, 31 Oct 2024 12:28:42 -0400
Subject: [PATCH] Adds heartbeat metric

---
 core/services/workflows/engine.go     |  1 +
 core/services/workflows/monitoring.go | 11 +++++++++++
 2 files changed, 12 insertions(+)

diff --git a/core/services/workflows/engine.go b/core/services/workflows/engine.go
index 8768d681375..7550d06cf95 100644
--- a/core/services/workflows/engine.go
+++ b/core/services/workflows/engine.go
@@ -1049,6 +1049,7 @@ func (e *Engine) heartbeat(ctx context.Context) {
 			e.logger.Info("shutting down heartbeat")
 			return
 		case <-ticker.C:
+			e.metrics.incrementEngineHeartbeatCounter(ctx)
 			logCustMsg(e.cma, "engine heartbeat at: "+e.clock.Now().Format(time.RFC3339), e.logger)
 		}
 	}
diff --git a/core/services/workflows/monitoring.go b/core/services/workflows/monitoring.go
index 8f5601554ed..bd448afd9e5 100644
--- a/core/services/workflows/monitoring.go
+++ b/core/services/workflows/monitoring.go
@@ -17,6 +17,7 @@ var workflowsRunningGauge metric.Int64Gauge
 var capabilityInvocationCounter metric.Int64Counter
 var workflowExecutionLatencyGauge metric.Int64Gauge // ms
 var workflowStepErrorCounter metric.Int64Counter
+var engineHeartbeatCounter metric.Int64UpDownCounter
 
 func initMonitoringResources() (err error) {
 	registerTriggerFailureCounter, err = beholder.GetMeter().Int64Counter("RegisterTriggerFailure")
@@ -44,6 +45,11 @@ func initMonitoringResources() (err error) {
 		return fmt.Errorf("failed to register workflow step error counter: %w", err)
 	}
 
+	engineHeartbeatCounter, err = beholder.GetMeter().Int64UpDownCounter("EngineHeartbeat")
+	if err != nil {
+		return fmt.Errorf("failed to register engine heartbeat counter: %w", err)
+	}
+
 	return nil
 }
 
@@ -82,6 +88,11 @@ func (c workflowsMetricLabeler) updateTotalWorkflowsGauge(ctx context.Context, v
 	workflowsRunningGauge.Record(ctx, val, metric.WithAttributes(otelLabels...))
 }
 
+func (c workflowsMetricLabeler) incrementEngineHeartbeatCounter(ctx context.Context) {
+	otelLabels := localMonitoring.KvMapToOtelAttributes(c.Labels)
+	engineHeartbeatCounter.Add(ctx, 1, metric.WithAttributes(otelLabels...))
+}
+
 // Observability keys
 const (
 	cIDKey  = "capabilityID"