Delete prometheus_rule_group metrics when groups are removed (#6693)

* Delete prometheus_rule_group metrics when groups are removed

Signed-off-by: Julien Pivotto <roidelapluie@inuits.eu>

rules/fixtures/rules2.yaml

@@ -0,0 +1,5 @@
+groups:
+  - name: test_2
+    rules:
+    - record: test_2
+      expr: vector(2)

rules/manager.go

@@ -874,7 +874,7 @@ func (m *Manager) Update(interval time.Duration, files []string, externalLabels
 		// check if new group equals with the old group, if yes then skip it.
 		// If not equals, stop it and wait for it to finish the current iteration.
 		// Then copy it into the new group.
-		gn := groupKey(newg.name, newg.file)
+		gn := groupKey(newg.file, newg.name)
 		oldg, ok := m.groups[gn]
 		delete(m.groups, gn)
 
@@ -901,8 +901,13 @@ func (m *Manager) Update(interval time.Duration, files []string, externalLabels
 	}
 
 	// Stop remaining old groups.
-	for _, oldg := range m.groups {
+	for n, oldg := range m.groups {
 		oldg.stop()
+		if m := oldg.metrics; m != nil {
+			m.groupLastEvalTime.DeleteLabelValues(n)
+			m.groupLastDuration.DeleteLabelValues(n)
+			m.groupRules.DeleteLabelValues(n)
+		}
 	}
 
 	wg.Wait()
@@ -958,7 +963,7 @@ func (m *Manager) LoadGroups(
 				))
 			}
 
-			groups[groupKey(rg.Name, fn)] = NewGroup(rg.Name, fn, itv, rules, shouldRestore, m.opts)
+			groups[groupKey(fn, rg.Name)] = NewGroup(rg.Name, fn, itv, rules, shouldRestore, m.opts)
 		}
 	}
 
@@ -966,8 +971,8 @@ func (m *Manager) LoadGroups(
 }
 
 // Group names need not be unique across filenames.
-func groupKey(name, file string) string {
+func groupKey(file, name string) string {
-	return name + ";" + file
+	return file + ";" + name
 }
 
 // RuleGroups returns the list of manager's rule groups.

rules/manager_test.go

@@ -24,6 +24,7 @@ import (
 	"time"
 
 	"github.com/go-kit/kit/log"
+	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/common/model"
 	yaml "gopkg.in/yaml.v2"
 
@@ -874,3 +875,80 @@ func TestNotify(t *testing.T) {
 	group.Eval(ctx, time.Unix(6, 0))
 	testutil.Equals(t, 1, len(lastNotified))
 }
+
+func TestMetricsUpdate(t *testing.T) {
+	files := []string{"fixtures/rules.yaml", "fixtures/rules2.yaml"}
+	metricNames := []string{
+		"prometheus_rule_group_interval_seconds",
+		"prometheus_rule_group_last_duration_seconds",
+		"prometheus_rule_group_last_evaluation_timestamp_seconds",
+		"prometheus_rule_group_rules",
+	}
+
+	storage := teststorage.New(t)
+	registry := prometheus.NewRegistry()
+	defer storage.Close()
+	opts := promql.EngineOpts{
+		Logger:        nil,
+		Reg:           nil,
+		MaxConcurrent: 10,
+		MaxSamples:    10,
+		Timeout:       10 * time.Second,
+	}
+	engine := promql.NewEngine(opts)
+	ruleManager := NewManager(&ManagerOptions{
+		Appendable: storage,
+		TSDB:       storage,
+		QueryFunc:  EngineQueryFunc(engine, storage),
+		Context:    context.Background(),
+		Logger:     log.NewNopLogger(),
+		Registerer: registry,
+	})
+	ruleManager.Run()
+	defer ruleManager.Stop()
+
+	countMetrics := func() int {
+		ms, err := registry.Gather()
+		testutil.Ok(t, err)
+		var metrics int
+		for _, m := range ms {
+			s := m.GetName()
+			for _, n := range metricNames {
+				if s == n {
+					metrics += len(m.Metric)
+					break
+				}
+			}
+		}
+		return metrics
+	}
+
+	cases := []struct {
+		files   []string
+		metrics int
+	}{
+		{
+			files:   files,
+			metrics: 8,
+		},
+		{
+			files:   files[:1],
+			metrics: 4,
+		},
+		{
+			files:   files[:0],
+			metrics: 0,
+		},
+		{
+			files:   files[1:],
+			metrics: 4,
+		},
+	}
+
+	for i, c := range cases {
+		err := ruleManager.Update(time.Second, c.files, nil)
+		testutil.Ok(t, err)
+		time.Sleep(2 * time.Second)
+		testutil.Equals(t, c.metrics, countMetrics(), "test %d: invalid count of metrics", i)
+	}
+}