From a16dc6b73eaf734843eeba8086b8aab33b963aa9 Mon Sep 17 00:00:00 2001
From: Rahix <rahix@rahix.de>
Date: Sun, 6 Oct 2024 11:29:04 +0200
Subject: [PATCH] succd: Add metrics for RP and RP operating time

Add counter metrics that count the total operating time for the roughing
pump and the diffusion pump.
---
 succbone/succd/http.go    |  8 ++++++++
 succbone/succd/process.go | 23 +++++++++++++++++++----
 2 files changed, 27 insertions(+), 4 deletions(-)

diff --git a/succbone/succd/http.go b/succbone/succd/http.go
index 07cf149..3445ec5 100644
--- a/succbone/succd/http.go
+++ b/succbone/succd/http.go
@@ -217,6 +217,14 @@ func (s *webServer) viewMetrics(w http.ResponseWriter, r *http.Request) {
 	fmt.Fprintf(w, "# HELP sem_vacuum_high_reached Whether the high vacuum threshold has been passed (boolean)\n")
 	fmt.Fprintf(w, "# TYPE sem_vacuum_high_reached gauge\n")
 	fmt.Fprintf(w, "sem_vacuum_high_reached %f\n", boolToFloat(high))
+
+	fmt.Fprintf(w, "# HELP sem_rp_operating_seconds_total Operating time of the roughing pump, in seconds\n")
+	fmt.Fprintf(w, "# TYPE sem_rp_operating_seconds_total counter\n")
+	fmt.Fprintf(w, "sem_rp_operating_seconds_total %f\n", state.rpOperatingTime.Seconds())
+
+	fmt.Fprintf(w, "# HELP sem_dp_operating_seconds_total Operating time of the diffusion pump, in seconds\n")
+	fmt.Fprintf(w, "# TYPE sem_dp_operating_seconds_total counter\n")
+	fmt.Fprintf(w, "sem_dp_operating_seconds_total %f\n", state.dpOperatingTime.Seconds())
 }
 
 func (s *webServer) viewRoughingPumpEnable(w http.ResponseWriter, r *http.Request) {
diff --git a/succbone/succd/process.go b/succbone/succd/process.go
index de99e84..cf0e5eb 100644
--- a/succbone/succd/process.go
+++ b/succbone/succd/process.go
@@ -51,6 +51,9 @@ type daemonState struct {
 	rpOn bool
 	dpOn bool
 
+	rpOperatingTime time.Duration
+	dpOperatingTime time.Duration
+
 	vent       momentaryOutput
 	pumpdown   momentaryOutput
 	aboveRough thresholdOutput
@@ -77,11 +80,15 @@ func (d *daemon) process(ctx context.Context) {
 		select {
 		case <-ticker.C:
 			now := time.Now()
-			if elapsed := now.Sub(lastRun); !lastRun.IsZero() && elapsed > periodGrace {
-				klog.Warningf("Processing loop lag: took %s, want %s", elapsed, period)
+			var elapsed time.Duration = 0
+			if !lastRun.IsZero() {
+				elapsed = now.Sub(lastRun)
+				if elapsed > periodGrace {
+					klog.Warningf("Processing loop lag: took %s, want %s", elapsed, period)
+				}
 			}
 			lastRun = now
-			if err := d.processOnce(ctx); err != nil {
+			if err := d.processOnce(ctx, elapsed); err != nil {
 				if errors.Is(err, ctx.Err()) {
 					return
 				} else {
@@ -99,7 +106,7 @@ func (d *daemon) process(ctx context.Context) {
 }
 
 // processOnce runs the main loop step of succd.
-func (d *daemon) processOnce(_ context.Context) error {
+func (d *daemon) processOnce(_ context.Context, elapsed time.Duration) error {
 	v, err := d.adcPirani.Read()
 	if err != nil {
 		return fmt.Errorf("when reading ADC: %w", err)
@@ -190,5 +197,13 @@ func (d *daemon) processOnce(_ context.Context) error {
 		}
 	}
 
+	// Update operating time counters
+	if d.rpOn && elapsed > 0 {
+		d.rpOperatingTime += elapsed
+	}
+	if d.dpOn && elapsed > 0 {
+		d.dpOperatingTime += elapsed
+	}
+
 	return nil
 }