chore: improvements

dannykopping · dannykopping · commit fcbfb7fed2dc · 2025-05-13T14:29:37.000+02:00
Signed-off-by: Danny Kopping &lt;dannykopping@gmail.com&gt;
diff --git a/enterprise/coderd/prebuilds/metricscollector.go b/enterprise/coderd/prebuilds/metricscollector.go
@@ -2,6 +2,7 @@ package prebuilds
 
 import (
 	"context"
+	"fmt"
 	"sync/atomic"
 	"time"
 
@@ -11,7 +12,6 @@ import (
 	"cdr.dev/slog"
 
 	"github.com/coder/coder/v2/coderd/database"
-	"github.com/coder/coder/v2/coderd/database/dbauthz"
 	"github.com/coder/coder/v2/coderd/prebuilds"
 )
 
@@ -57,6 +57,12 @@ var (
 		labels,
 		nil,
 	)
+	lastUpdateDesc = prometheus.NewDesc(
+		"coderd_prebuilt_workspaces_metrics_last_updated",
+		"The unix timestamp when the metrics related to prebuilt workspaces were last updated; these metrics are cached.",
+		[]string{},
+		nil,
+	)
 )
 
 const (
@@ -74,7 +80,6 @@ type MetricsCollector struct {
 
 var _ prometheus.Collector = new(MetricsCollector)
 
-// NewMetricsCollector returns a
 func NewMetricsCollector(db database.Store, logger slog.Logger, snapshotter prebuilds.StateSnapshotter) *MetricsCollector {
 	log := logger.Named("prebuilds_metrics_collector")
 	return &MetricsCollector{
@@ -91,18 +96,16 @@ func (*MetricsCollector) Describe(descCh chan<- *prometheus.Desc) {
 	descCh <- desiredPrebuildsDesc
 	descCh <- runningPrebuildsDesc
 	descCh <- eligiblePrebuildsDesc
+	descCh <- lastUpdateDesc
 }
 
 // Collect uses the cached state to set configured metrics.
 // The state is cached because this function can be called multiple times per second and retrieving the current state
 // is an expensive operation.
 func (mc *MetricsCollector) Collect(metricsCh chan<- prometheus.Metric) {
-	// nolint:gocritic // We need to set an authz context to read metrics from the db.
-	ctx := dbauthz.AsPrebuildsOrchestrator(context.Background())
-
-	currentState := mc.latestState.Load()
+	currentState := mc.latestState.Load() // Grab a copy; it's ok if it goes stale during the course of this func.
 	if currentState == nil {
-		mc.logger.Warn(ctx, "failed to set prebuilds metrics; state not set")
+		mc.logger.Warn(context.Background(), "failed to set prebuilds metrics; state not set")
 		return
 	}
 
@@ -119,7 +122,7 @@ func (mc *MetricsCollector) Collect(metricsCh chan<- prometheus.Metric) {
 
 		presetSnapshot, err := currentState.snapshot.FilterByPreset(preset.ID)
 		if err != nil {
-			mc.logger.Error(ctx, "failed to filter by preset", slog.Error(err))
+			mc.logger.Error(context.Background(), "failed to filter by preset", slog.Error(err))
 			continue
 		}
 		state := presetSnapshot.CalculateState()
@@ -128,11 +131,14 @@ func (mc *MetricsCollector) Collect(metricsCh chan<- prometheus.Metric) {
 		metricsCh <- prometheus.MustNewConstMetric(runningPrebuildsDesc, prometheus.GaugeValue, float64(state.Actual), preset.TemplateName, preset.Name, preset.OrganizationName)
 		metricsCh <- prometheus.MustNewConstMetric(eligiblePrebuildsDesc, prometheus.GaugeValue, float64(state.Eligible), preset.TemplateName, preset.Name, preset.OrganizationName)
 	}
+
+	metricsCh <- prometheus.MustNewConstMetric(lastUpdateDesc, prometheus.GaugeValue, float64(currentState.createdAt.Unix()))
 }
 
 type state struct {
 	prebuildMetrics []database.GetPrebuildMetricsRow
 	snapshot        *prebuilds.GlobalSnapshot
+	createdAt       time.Time
 }
 
 // BackgroundFetch updates the metrics state every given interval.
@@ -157,6 +163,7 @@ func (mc *MetricsCollector) BackgroundFetch(ctx context.Context, updateInterval,
 
 // UpdateState builds the current metrics state.
 func (mc *MetricsCollector) UpdateState(ctx context.Context, timeout time.Duration) error {
+	start := time.Now()
 	mc.logger.Debug(ctx, "fetching prebuilds metrics state")
 	fetchCtx, fetchCancel := context.WithTimeout(ctx, timeout)
 	defer fetchCancel()
@@ -170,11 +177,12 @@ func (mc *MetricsCollector) UpdateState(ctx context.Context, timeout time.Durati
 	if err != nil {
 		return xerrors.Errorf("snapshot state: %w", err)
 	}
-	mc.logger.Debug(ctx, "fetched prebuilds metrics state")
+	mc.logger.Debug(ctx, "fetched prebuilds metrics state", slog.F("duration_secs", fmt.Sprintf("%.2f", time.Since(start).Seconds())))
 
 	mc.latestState.Store(&state{
 		prebuildMetrics: prebuildMetrics,
 		snapshot:        snapshot,
+		createdAt:       time.Now(),
 	})
 	return nil
 }
diff --git a/enterprise/coderd/prebuilds/reconcile.go b/enterprise/coderd/prebuilds/reconcile.go
@@ -5,6 +5,7 @@ import (
 	"database/sql"
 	"fmt"
 	"math"
+	"sync"
 	"sync/atomic"
 	"time"
 
@@ -87,10 +88,12 @@ func (c *StoreReconciler) Run(ctx context.Context) {
 		slog.F("backoff_interval", c.cfg.ReconciliationBackoffInterval.String()),
 		slog.F("backoff_lookback", c.cfg.ReconciliationBackoffLookback.String()))
 
+	var wg sync.WaitGroup
 	ticker := c.clock.NewTicker(reconciliationInterval)
 	defer ticker.Stop()
 	defer func() {
 		c.done <- struct{}{}
+		wg.Wait()
 	}()
 
 	// nolint:gocritic // Reconciliation Loop needs Prebuilds Orchestrator permissions.
@@ -99,7 +102,11 @@ func (c *StoreReconciler) Run(ctx context.Context) {
 
 	// Start updating metrics in the background.
 	if c.metrics != nil {
-		go c.metrics.BackgroundFetch(ctx, metricsUpdateInterval, metricsUpdateTimeout)
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			c.metrics.BackgroundFetch(ctx, metricsUpdateInterval, metricsUpdateTimeout)
+		}()
 	}
 
 	// Everything is in place, reconciler can now be considered as running.
