Filter out unacceptable labels in AgentStats

dannykopping · dannykopping · commit 3c1d016a82d6 · 2024-03-11T16:09:13.000Z
Signed-off-by: Danny Kopping &lt;danny@coder.com&gt;
diff --git a/coderd/prometheusmetrics/aggregator_test.go b/coderd/prometheusmetrics/aggregator_test.go
@@ -262,7 +262,7 @@ func TestUpdateMetrics_MetricsExpire(t *testing.T) {
 
 	// given
 	registry := prometheus.NewRegistry()
-	metricsAggregator, err := prometheusmetrics.NewMetricsAggregator(slogtest.Make(t, &slogtest.Options{IgnoreErrors: true}), registry, time.Millisecond, nil)
+	metricsAggregator, err := prometheusmetrics.NewMetricsAggregator(slogtest.Make(t, &slogtest.Options{IgnoreErrors: true}), registry, time.Millisecond, agentmetrics.LabelAll)
 	require.NoError(t, err)
 
 	ctx, cancelFunc := context.WithCancel(context.Background())
diff --git a/coderd/prometheusmetrics/prometheusmetrics.go b/coderd/prometheusmetrics/prometheusmetrics.go
@@ -338,6 +338,8 @@ func AgentStats(ctx context.Context, logger slog.Logger, registerer prometheus.R
 		aggregateByLabels = agentmetrics.LabelAgentStats
 	}
 
+	aggregateByLabels = filterInvalidLabels(aggregateByLabels)
+
 	metricsCollectorAgentStats := prometheus.NewHistogram(prometheus.HistogramOpts{
 		Namespace: "coderd",
 		Subsystem: "prometheusmetrics",
@@ -514,3 +516,20 @@ func AgentStats(ctx context.Context, logger slog.Logger, registerer prometheus.R
 		<-done
 	}, nil
 }
+
+// filterInvalidLabels handles a slightly messy situation whereby `prometheus-aggregate-agent-stats-by` can control on
+// which labels agent stats are aggregated, but for these specific metrics in this file there is no `template` label value,
+// and therefore we have to exclude it from the list of acceptable labels.
+func filterInvalidLabels(labels []string) []string {
+	var out []string
+
+	for _, label := range labels {
+		if label == agentmetrics.LabelTemplateName {
+			continue
+		}
+
+		out = append(out, label)
+	}
+
+	return out
+}
diff --git a/coderd/prometheusmetrics/prometheusmetrics_test.go b/coderd/prometheusmetrics/prometheusmetrics_test.go
@@ -5,6 +5,7 @@ import (
 	"database/sql"
 	"encoding/json"
 	"fmt"
+	"github.com/coder/coder/v2/coderd/agentmetrics"
 	"os"
 	"reflect"
 	"sync/atomic"
@@ -451,7 +452,7 @@ func TestAgentStats(t *testing.T) {
 	// and it doesn't depend on the real time.
 	closeFunc, err := prometheusmetrics.AgentStats(ctx, slogtest.Make(t, &slogtest.Options{
 		IgnoreErrors: true,
-	}), registry, db, time.Now().Add(-time.Minute), time.Millisecond, nil)
+	}), registry, db, time.Now().Add(-time.Minute), time.Millisecond, agentmetrics.LabelAll)
 	require.NoError(t, err)
 	t.Cleanup(closeFunc)
 
