Add Badger expvar metrics to Prometheus metrics. (#5094)

Fixes #4772 

This adds the following badger metrics into /debug/prometheus_metrics:

badger_v2_disk_reads_total
badger_v2_disk_writes_total
badger_v2_gets_total
badger_v2_lsm_bloom_hits_total (per level)
badger_v2_lsm_level_gets_total (per level)
badger_v2_memtable_gets_total
badger_v2_puts_total
badger_v2_read_bytes
badger_v2_written_bytes

This is added via the Prometheus expvar collector.
Update metrics_test.go for the seven initial Badger versions (excluding the LSM metrics).

This adds to the exposed Prometheus metrics. These metrics were already accessible via /debug/vars.

The LSM metrics don't show up immediately. They show up after there are hits/gets to the LSM tree.

Changes
* Add Badger metrics to Prometheus
* Update metrics test to check for badger metrics.

Co-authored-by: Ibrahim Jarif <ibrahim@dgraph.io>

dgraph/cmd/alpha/metrics_test.go

@@ -38,13 +38,11 @@ func TestMetrics(t *testing.T) {
 		// Go Runtime Metrics
 		"go_goroutines", "go_memstats_gc_cpu_fraction", "go_memstats_heap_alloc_bytes",
 		"go_memstats_heap_idle_bytes", "go_memstats_heap_inuse_bytes", "dgraph_latency_bucket",
-		// TODO: add support for the following Badger metrics, which is currently only available
-		// through /debug/vars. Consider manually add them as OpenCensus metrics, and then
-		// test them here
 
-		// "badger_disk_reads_total", "badger_disk_writes_total", "badger_gets_total",
-		// "badger_memtable_gets_total", "badger_puts_total", "badger_read_bytes",
-		// "badger_written_bytes",
+		// Badger Metrics
+		"badger_v2_disk_reads_total", "badger_v2_disk_writes_total", "badger_v2_gets_total",
+		"badger_v2_memtable_gets_total", "badger_v2_puts_total", "badger_v2_read_bytes",
+		"badger_v2_written_bytes",
 
 		// Dgraph Memory Metrics
 		"dgraph_memory_idle_bytes", "dgraph_memory_inuse_bytes", "dgraph_memory_proc_bytes",
@@ -61,7 +59,7 @@ func TestMetrics(t *testing.T) {
 
 func extractMetrics(metrics string) (map[string]interface{}, error) {
 	lines := strings.Split(metrics, "\n")
-	metricRegex, err := regexp.Compile("(^[a-z_]+)")
+	metricRegex, err := regexp.Compile("(^[a-z0-9_]+)")
 	if err != nil {
 		return nil, err
 	}

x/metrics.go

@@ -223,7 +223,11 @@ func init() {
 
 	CheckfNoTrace(view.Register(allViews...))
 
+	prometheus.MustRegister(NewBadgerCollector())
+
 	pe, err := oc_prom.NewExporter(oc_prom.Options{
+		// DefaultRegisterer includes a ProcessCollector for process_* metrics, a GoCollector for
+		// go_* metrics, and the badger_* metrics.
 		Registry:  prometheus.DefaultRegisterer.(*prometheus.Registry),
 		Namespace: "dgraph",
 		OnError:   func(err error) { glog.Errorf("%v", err) },
@@ -234,6 +238,67 @@ func init() {
 	http.Handle("/debug/prometheus_metrics", pe)
 }
 
+// NewBadgerCollector returns a prometheus Collector for Badger metrics from expvar.
+func NewBadgerCollector() prometheus.Collector {
+	return prometheus.NewExpvarCollector(map[string]*prometheus.Desc{
+		"badger_v2_disk_reads_total": prometheus.NewDesc(
+			"badger_v2_disk_reads_total",
+			"Number of cumulative reads by Badger",
+			nil, nil,
+		),
+		"badger_v2_disk_writes_total": prometheus.NewDesc(
+			"badger_v2_disk_writes_total",
+			"Number of cumulative writes by Badger",
+			nil, nil,
+		),
+		"badger_v2_read_bytes": prometheus.NewDesc(
+			"badger_v2_read_bytes",
+			"Number of cumulative bytes read by Badger",
+			nil, nil,
+		),
+		"badger_v2_written_bytes": prometheus.NewDesc(
+			"badger_v2_written_bytes",
+			"Number of cumulative bytes written by Badger",
+			nil, nil,
+		),
+		"badger_v2_lsm_level_gets_total": prometheus.NewDesc(
+			"badger_v2_lsm_level_gets_total",
+			"Total number of LSM gets",
+			[]string{"level"}, nil,
+		),
+		"badger_v2_lsm_bloom_hits_total": prometheus.NewDesc(
+			"badger_v2_lsm_bloom_hits_total",
+			"Total number of LSM bloom hits",
+			[]string{"level"}, nil,
+		),
+		"badger_v2_gets_total": prometheus.NewDesc(
+			"badger_v2_gets_total",
+			"Total number of gets",
+			nil, nil,
+		),
+		"badger_v2_puts_total": prometheus.NewDesc(
+			"badger_v2_puts_total",
+			"Total number of puts",
+			nil, nil,
+		),
+		"badger_v2_memtable_gets_total": prometheus.NewDesc(
+			"badger_v2_memtable_gets_total",
+			"Total number of memtable gets",
+			nil, nil,
+		),
+		"badger_v2_lsm_size": prometheus.NewDesc(
+			"badger_v2_lsm_size",
+			"Size of the LSM in bytes",
+			[]string{"dir"}, nil,
+		),
+		"badger_v2_vlog_size": prometheus.NewDesc(
+			"badger_v2_vlog_size",
+			"Size of the value log in bytes",
+			[]string{"dir"}, nil,
+		),
+	})
+}
+
 // MetricsContext returns a context with tags that are useful for
 // distinguishing the state of the running system.
 // This context will be used to derive other contexts.