vmalert: add function "query", "first" and "value" to alert templates functions (#960)

The commit adds a support for template function `query`,
`first` and `value`. The function `query` executes
a MetricsQL query for active alerts. In vmalert we
update templates on every evaluation for active alerts
to keep them up to date. With `query` func it may become
a perf issue since it will fire a query on every execution.
We should keep it in mind for now.

https://github.com/VictoriaMetrics/VictoriaMetrics/issues/539

app/vmalert/README.md

@@ -21,7 +21,6 @@ may fail;
 * by default, rules execution is sequential within one group, but persisting of execution results to remote
 storage is asynchronous. Hence, user shouldn't rely on recording rules chaining when result of previous
 recording rule is reused in next one;
-* there is no `query` function support in templates yet;
 * `vmalert` has no UI, just an API for getting groups and rules statuses.
 
 ### QuickStart

app/vmalert/alerting.go

@@ -137,6 +137,7 @@ func (ar *AlertingRule) Exec(ctx context.Context, q datasource.Querier, series b
 		}
 	}
 
+	qFn := func(query string) ([]datasource.Metric, error) { return q.Query(ctx, query) }
 	updated := make(map[uint64]struct{})
 	// update list of active alerts
 	for _, m := range qMetrics {
@@ -158,14 +159,14 @@ func (ar *AlertingRule) Exec(ctx context.Context, q datasource.Querier, series b
 				a.Value = m.Value
 				// and re-exec template since Value can be used
 				// in templates
-				err = ar.template(a)
+				err = ar.template(a, qFn)
 				if err != nil {
 					return nil, err
 				}
 			}
 			continue
 		}
-		a, err := ar.newAlert(m, ar.lastExecTime)
+		a, err := ar.newAlert(m, ar.lastExecTime, qFn)
 		if err != nil {
 			ar.lastExecError = err
 			return nil, fmt.Errorf("failed to create alert: %w", err)
@@ -245,7 +246,7 @@ func hash(m datasource.Metric) uint64 {
 	return hash.Sum64()
 }
 
-func (ar *AlertingRule) newAlert(m datasource.Metric, start time.Time) (*notifier.Alert, error) {
+func (ar *AlertingRule) newAlert(m datasource.Metric, start time.Time, qFn notifier.QueryFn) (*notifier.Alert, error) {
 	a := &notifier.Alert{
 		GroupID: ar.GroupID,
 		Name:    ar.Name,
@@ -264,16 +265,16 @@ func (ar *AlertingRule) newAlert(m datasource.Metric, start time.Time) (*notifie
 		}
 		a.Labels[l.Name] = l.Value
 	}
-	return a, ar.template(a)
+	return a, ar.template(a, qFn)
 }
 
-func (ar *AlertingRule) template(a *notifier.Alert) error {
+func (ar *AlertingRule) template(a *notifier.Alert, qFn notifier.QueryFn) error {
 	var err error
-	a.Labels, err = a.ExecTemplate(a.Labels)
+	a.Labels, err = a.ExecTemplate(qFn, a.Labels)
 	if err != nil {
 		return err
 	}
-	a.Annotations, err = a.ExecTemplate(ar.Annotations)
+	a.Annotations, err = a.ExecTemplate(qFn, ar.Annotations)
 	return err
 }
 
@@ -393,6 +394,8 @@ func (ar *AlertingRule) Restore(ctx context.Context, q datasource.Querier, lookb
 		return fmt.Errorf("querier is nil")
 	}
 
+	qFn := func(query string) ([]datasource.Metric, error) { return q.Query(ctx, query) }
+
 	// account for external labels in filter
 	var labelsFilter string
 	for k, v := range labels {
@@ -421,7 +424,7 @@ func (ar *AlertingRule) Restore(ctx context.Context, q datasource.Querier, lookb
 			m.Labels = append(m.Labels, l)
 		}
 
-		a, err := ar.newAlert(m, time.Unix(int64(m.Value), 0))
+		a, err := ar.newAlert(m, time.Unix(int64(m.Value), 0), qFn)
 		if err != nil {
 			return fmt.Errorf("failed to create alert: %w", err)
 		}

app/vmalert/config/config_test.go

@@ -7,8 +7,9 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
 	"gopkg.in/yaml.v2"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
 )
 
 func TestMain(m *testing.M) {
@@ -42,7 +43,7 @@ func TestParseBad(t *testing.T) {
 		},
 		{
 			[]string{"testdata/dir/rules2-bad.rules"},
-			"function \"value\" not defined",
+			"function \"unknown\" not defined",
 		},
 		{
 			[]string{"testdata/dir/rules3-bad.rules"},
@@ -137,12 +138,14 @@ func TestGroup_Validate(t *testing.T) {
 						Alert: "alert",
 						Expr:  "up == 1",
 						Labels: map[string]string{
-							"summary": "{{ value|query }}",
+							"summary": `
+{{ with printf "node_memory_MemTotal{job='node',instance='%s'}" "localhost" | query }}
+  {{ . | first | value | humanize1024 }}B
+{{ end }}`,
 						},
 					},
 				},
 			},
-			expErr:              "error parsing annotation",
 			validateAnnotations: true,
 		},
 		{

app/vmalert/config/testdata/dir/rules2-bad.rules

@@ -6,6 +6,6 @@ groups:
         expr: vm_rows > 0
         labels:
           label: bar
-          summary: "{{ value|query }}"
+          summary: "{{ unknown|query }}"
         annotations:
           description: "{{$labels}}"

app/vmalert/config/testdata/rules2-good.rules

@@ -7,11 +7,18 @@ groups:
         expr: sum(vm_tcplistener_conns) by(instance) > 1
         for: 3m
         annotations:
-          summary: "Too high connection number for {{$labels.instance}}"
+          summary: Too high connection number for {{$labels.instance}}
+            {{ with printf "sum(vm_tcplistener_conns{instance=%q})" .Labels.instance | query }}
+              {{ . | first | value }}
+            {{ end }}
           description: "It is {{ $value }} connections for {{$labels.instance}}"
       - alert: ExampleAlertAlwaysFiring
         expr: sum by(job)
           (up == 1)
+        annotations:
+          summary: Instances up {{ range query "up" }}
+            {{ . | label "instance" }}
+            {{ end }}
       - record: handler:requests:rate5m
         expr: sum(rate(prometheus_http_requests_total[5m])) by (handler)
         labels:

app/vmalert/datasource/datasource.go

@@ -34,6 +34,17 @@ func (m *Metric) AddLabel(key, value string) {
 	m.Labels = append(m.Labels, Label{Name: key, Value: value})
 }
 
+// Label returns the given label value.
+// If label is missing empty string will be returned
+func (m *Metric) Label(key string) string {
+	for _, l := range m.Labels {
+		if l.Name == key {
+			return l.Value
+		}
+	}
+	return ""
+}
+
 // Label represents metric's label
 type Label struct {
 	Name  string

app/vmalert/group_test.go

@@ -167,7 +167,7 @@ func TestGroupStart(t *testing.T) {
 	m2 := metricWithLabels(t, "instance", inst2, "job", job)
 
 	r := g.Rules[0].(*AlertingRule)
-	alert1, err := r.newAlert(m1, time.Now())
+	alert1, err := r.newAlert(m1, time.Now(), nil)
 	if err != nil {
 		t.Fatalf("faield to create alert: %s", err)
 	}
@@ -179,7 +179,7 @@ func TestGroupStart(t *testing.T) {
 	alert1.Labels["host"] = inst1
 	alert1.ID = hash(m1)
 
-	alert2, err := r.newAlert(m2, time.Now())
+	alert2, err := r.newAlert(m2, time.Now(), nil)
 	if err != nil {
 		t.Fatalf("faield to create alert: %s", err)
 	}

app/vmalert/main.go

@@ -206,7 +206,7 @@ func getAlertURLGenerator(externalURL *url.URL, externalAlertSource string, vali
 		"tpl": externalAlertSource,
 	}
 	return func(alert notifier.Alert) string {
-		templated, err := alert.ExecTemplate(m)
+		templated, err := alert.ExecTemplate(nil, m)
 		if err != nil {
 			logger.Errorf("can not exec source template %s", err)
 		}

app/vmalert/notifier/alert.go

@@ -62,21 +62,23 @@ const tplHeader = `{{ $value := .Value }}{{ $labels := .Labels }}{{ $expr := .Ex
 
 // ExecTemplate executes the Alert template for give
 // map of annotations.
-func (a *Alert) ExecTemplate(annotations map[string]string) (map[string]string, error) {
+// Every alert could have a different datasource, so function
+// requires a queryFunction as an argument.
+func (a *Alert) ExecTemplate(q QueryFn, annotations map[string]string) (map[string]string, error) {
 	tplData := alertTplData{Value: a.Value, Labels: a.Labels, Expr: a.Expr}
-	return templateAnnotations(annotations, tplHeader, tplData)
+	return templateAnnotations(annotations, tplData, funcsWithQuery(q))
 }
 
 // ValidateTemplates validate annotations for possible template error, uses empty data for template population
 func ValidateTemplates(annotations map[string]string) error {
-	_, err := templateAnnotations(annotations, tplHeader, alertTplData{
+	_, err := templateAnnotations(annotations, alertTplData{
 		Labels: map[string]string{},
 		Value:  0,
-	})
+	}, tmplFunc)
 	return err
 }
 
-func templateAnnotations(annotations map[string]string, header string, data alertTplData) (map[string]string, error) {
+func templateAnnotations(annotations map[string]string, data alertTplData, funcs template.FuncMap) (map[string]string, error) {
 	var builder strings.Builder
 	var buf bytes.Buffer
 	eg := new(utils.ErrGroup)
@@ -85,10 +87,10 @@ func templateAnnotations(annotations map[string]string, header string, data aler
 		r[key] = text
 		buf.Reset()
 		builder.Reset()
-		builder.Grow(len(header) + len(text))
-		builder.WriteString(header)
+		builder.Grow(len(tplHeader) + len(text))
+		builder.WriteString(tplHeader)
 		builder.WriteString(text)
-		if err := templateAnnotation(&buf, builder.String(), data); err != nil {
+		if err := templateAnnotation(&buf, builder.String(), data, funcs); err != nil {
 			eg.Add(fmt.Errorf("key %q, template %q: %w", key, text, err))
 			continue
 		}
@@ -97,8 +99,9 @@ func templateAnnotations(annotations map[string]string, header string, data aler
 	return r, eg.Err()
 }
 
-func templateAnnotation(dst io.Writer, text string, data alertTplData) error {
-	tpl, err := template.New("").Funcs(tmplFunc).Option("missingkey=zero").Parse(text)
+func templateAnnotation(dst io.Writer, text string, data alertTplData, funcs template.FuncMap) error {
+	t := template.New("").Funcs(funcs).Option("missingkey=zero")
+	tpl, err := t.Parse(text)
 	if err != nil {
 		return fmt.Errorf("error parsing annotation: %w", err)
 	}

app/vmalert/notifier/alert_test.go

@@ -2,6 +2,8 @@ package notifier
 
 import (
 	"testing"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 )
 
 func TestAlert_ExecTemplate(t *testing.T) {
@@ -60,11 +62,41 @@ func TestAlert_ExecTemplate(t *testing.T) {
 				"exprEscapedPath":  "vm_rows%7B%5C%22label%5C%22=%5C%22bar%5C%22%7D%3E0",
 			},
 		},
+		{
+			name:  "query",
+			alert: &Alert{Expr: `vm_rows{"label"="bar"}>0`},
+			annotations: map[string]string{
+				"summary": `{{ query "foo" | first | value }}`,
+				"desc":    `{{ range query "bar" }}{{ . | label "foo" }} {{ . | value }};{{ end }}`,
+			},
+			expTpl: map[string]string{
+				"summary": "1",
+				"desc":    "bar 1;garply 2;",
+			},
+		},
 	}
 
+	qFn := func(q string) ([]datasource.Metric, error) {
+		return []datasource.Metric{
+			{
+				Labels: []datasource.Label{
+					{Name: "foo", Value: "bar"},
+					{Name: "baz", Value: "qux"},
+				},
+				Value: 1,
+			},
+			{
+				Labels: []datasource.Label{
+					{Name: "foo", Value: "garply"},
+					{Name: "baz", Value: "fred"},
+				},
+				Value: 2,
+			},
+		}, nil
+	}
 	for _, tc := range testCases {
 		t.Run(tc.name, func(t *testing.T) {
-			tpl, err := tc.alert.ExecTemplate(tc.annotations)
+			tpl, err := tc.alert.ExecTemplate(qFn, tc.annotations)
 			if err != nil {
 				t.Fatal(err)
 			}

app/vmalert/notifier/template_func.go

@@ -14,21 +14,40 @@
 package notifier
 
 import (
+	"errors"
 	"fmt"
-	html_template "html/template"
 	"math"
 	"net/url"
 	"regexp"
 	"strings"
-	text_template "text/template"
 	"time"
+
+	htmlTpl "html/template"
+	textTpl "text/template"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 )
 
-var tmplFunc text_template.FuncMap
+// QueryFn is used to wrap a call to datasource into simple-to-use function
+// for templating functions.
+type QueryFn func(query string) ([]datasource.Metric, error)
 
-// InitTemplateFunc returns template helper functions
+func funcsWithQuery(query QueryFn) textTpl.FuncMap {
+	fm := make(textTpl.FuncMap)
+	for k, fn := range tmplFunc {
+		fm[k] = fn
+	}
+	fm["query"] = func(q string) ([]datasource.Metric, error) {
+		return query(q)
+	}
+	return fm
+}
+
+var tmplFunc textTpl.FuncMap
+
+// InitTemplateFunc initiates template helper functions
 func InitTemplateFunc(externalURL *url.URL) {
-	tmplFunc = text_template.FuncMap{
+	tmplFunc = textTpl.FuncMap{
 		"args": func(args ...interface{}) map[string]interface{} {
 			result := make(map[string]interface{})
 			for i, a := range args {
@@ -40,8 +59,8 @@ func InitTemplateFunc(externalURL *url.URL) {
 			re := regexp.MustCompile(pattern)
 			return re.ReplaceAllString(text, repl)
 		},
-		"safeHtml": func(text string) html_template.HTML {
-			return html_template.HTML(text)
+		"safeHtml": func(text string) htmlTpl.HTML {
+			return htmlTpl.HTML(text)
 		},
 		"match":   regexp.MatchString,
 		"title":   strings.Title,
@@ -151,6 +170,24 @@ func InitTemplateFunc(externalURL *url.URL) {
 		"quotesEscape": func(q string) string {
 			return strings.Replace(q, `"`, `\"`, -1)
 		},
+		// query function supposed to be substituted at funcsWithQuery().
+		// it is present here only for validation purposes, when there is no
+		// provided datasource.
+		"query": func(q string) ([]datasource.Metric, error) {
+			return nil, nil
+		},
+		"first": func(metrics []datasource.Metric) (datasource.Metric, error) {
+			if len(metrics) > 0 {
+				return metrics[0], nil
+			}
+			return datasource.Metric{}, errors.New("first() called on vector with no elements")
+		},
+		"label": func(label string, m datasource.Metric) string {
+			return m.Label(label)
+		},
+		"value": func(m datasource.Metric) float64 {
+			return m.Value
+		},
 	}
 }