Add metrics for cache hits/misses (#375)

* Add metrics for cache hits/misses

Add counters for cache middleware hits and misses.

* Add test for cache middleware hit/miss counters.

* Fix cache hit metric incrementing.

* Add cache hit/miss metrics to dnssec middleware.

* Update README metric documentation.

middleware/cache/README.md

@@ -39,8 +39,12 @@ The minimum TTL allowed on resource records is 5 seconds.
 
 If monitoring is enabled (via the *prometheus* directive) then the following metrics are exported:
 
-* coredns_cache_size{type} - total elements in the case, type is either "denial" or "success".
-* coredns_cache_capacity{type} - total capacity of the cache, type is either "denial" or "success".
+* coredns_cache_size{type} - Total elements in the cache by cache type.
+* coredns_cache_capacity{type} - Total capacity of the cache by cache type.
+* coredns_cache_hits_total{type} - Counter of cache hits by cache type.
+* coredns_cache_misses_total - Counter of cache misses.
+
+Cache types are either "denial" or "success".
 
 ## Examples
 

middleware/cache/handler.go

@@ -44,12 +44,15 @@ func (c *Cache) get(qname string, qtype uint16, do bool) (*item, bool, bool) {
 	k := rawKey(qname, qtype, do)
 
 	if i, ok := c.ncache.Get(k); ok {
+		cacheHits.WithLabelValues(Denial).Inc()
 		return i.(*item), ok, i.(*item).expired(time.Now())
 	}
 
 	if i, ok := c.pcache.Get(k); ok {
+		cacheHits.WithLabelValues(Success).Inc()
 		return i.(*item), ok, i.(*item).expired(time.Now())
 	}
+	cacheMisses.Inc()
 	return nil, false, false
 }
 
@@ -67,6 +70,20 @@ var (
 		Name:      "capacity",
 		Help:      "The cache's capacity.",
 	}, []string{"type"})
+
+	cacheHits = prometheus.NewCounterVec(prometheus.CounterOpts{
+		Namespace: middleware.Namespace,
+		Subsystem: subsystem,
+		Name:      "hits_total",
+		Help:      "The count of cache hits.",
+	}, []string{"type"})
+
+	cacheMisses = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: middleware.Namespace,
+		Subsystem: subsystem,
+		Name:      "misses_total",
+		Help:      "The count of cache misses.",
+	})
 )
 
 const subsystem = "cache"
@@ -74,4 +91,6 @@ const subsystem = "cache"
 func init() {
 	prometheus.MustRegister(cacheSize)
 	prometheus.MustRegister(cacheCapacity)
+	prometheus.MustRegister(cacheHits)
+	prometheus.MustRegister(cacheMisses)
 }

middleware/dnssec/README.md

@@ -43,5 +43,7 @@ If monitoring is enabled (via the *prometheus* directive) then the following met
 
 * coredns_dnssec_cache_size{type} - total elements in the cache, type is "signature".
 * coredns_dnssec_cache_capacity{type} - total capacity of the cache, type is "signature".
+* coredns_dnssec_cache_hits_total - Counter of cache hits.
+* coredns_dnssec_cache_misses_total - Counter of cache misses.
 
 ## Examples

middleware/dnssec/dnssec.go

@@ -115,8 +115,10 @@ func (d Dnssec) set(key string, sigs []dns.RR) {
 
 func (d Dnssec) get(key string) ([]dns.RR, bool) {
 	if s, ok := d.cache.Get(key); ok {
+		cacheHits.Inc()
 		return s.([]dns.RR), true
 	}
+	cacheMisses.Inc()
 	return nil, false
 }
 

middleware/dnssec/handler.go

@@ -53,6 +53,20 @@ var (
 		Name:      "cache_capacity",
 		Help:      "The dnssec cache's capacity.",
 	}, []string{"type"})
+
+	cacheHits = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: middleware.Namespace,
+		Subsystem: subsystem,
+		Name:      "cache_hits_total",
+		Help:      "The count of cache hits.",
+	})
+
+	cacheMisses = prometheus.NewCounter(prometheus.CounterOpts{
+		Namespace: middleware.Namespace,
+		Subsystem: subsystem,
+		Name:      "cache_misses_total",
+		Help:      "The count of cache misses.",
+	})
 )
 
 // Name implements the Handler interface.
@@ -63,4 +77,6 @@ const subsystem = "dnssec"
 func init() {
 	prometheus.MustRegister(cacheSize)
 	prometheus.MustRegister(cacheCapacity)
+	prometheus.MustRegister(cacheHits)
+	prometheus.MustRegister(cacheMisses)
 }

test/metrics_test.go

@@ -73,7 +73,8 @@ func TestMetricsRefused(t *testing.T) {
 }
 
 func TestMetricsCache(t *testing.T) {
-	metricName := "coredns_cache_size"
+	cacheSizeMetricName := "coredns_cache_size"
+	cacheHitMetricName := "coredns_cache_hits_total"
 
 	corefile := `example.net:0 {
 	proxy . 8.8.8.8:53
@@ -97,11 +98,24 @@ func TestMetricsCache(t *testing.T) {
 	}
 
 	data := mtest.Scrape(t, "http://"+metrics.ListenAddr+"/metrics")
-	// Get the value for the metrics where the one of the labels values matches "success".
-	got, _ := mtest.MetricValueLabel(metricName, cache.Success, data)
+	// Get the value for the cache size metric where the one of the labels values matches "success".
+	got, _ := mtest.MetricValueLabel(cacheSizeMetricName, cache.Success, data)
 
 	if got != "1" {
-		t.Errorf("Expected value %s for %s, but got %s", "1", metricName, got)
+		t.Errorf("Expected value %s for %s, but got %s", "1", cacheSizeMetricName, got)
+	}
+
+	// Second request for the same response to test hit counter.
+	if _, err = dns.Exchange(m, udp); err != nil {
+		t.Fatalf("Could not send message: %s", err)
+	}
+
+	data = mtest.Scrape(t, "http://"+metrics.ListenAddr+"/metrics")
+	// Get the value for the cache hit counter where the one of the labels values matches "success".
+	got, _ = mtest.MetricValueLabel(cacheHitMetricName, cache.Success, data)
+
+	if got != "1" {
+		t.Errorf("Expected value %s for %s, but got %s", "1", cacheHitMetricName, got)
 	}
 }