func extractCounter(out Ingester, o *ProcessOptions, f *dto.MetricFamily) error {
samples := make(model.Samples, 0, len(f.Metric))
for _, m := range f.Metric {
if m.Counter == nil {
continue
}
sample := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(m.Counter.GetValue()),
}
samples = append(samples, sample)
if m.TimestampMs != nil {
sample.Timestamp = model.TimestampFromUnixNano(*m.TimestampMs * 1000000)
} else {
sample.Timestamp = o.Timestamp
}
metric := sample.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.MetricNameLabel] = model.LabelValue(f.GetName())
}
return out.Ingest(samples)
}
func parseTime(s string) (clientmodel.Timestamp, error) {
if t, err := strconv.ParseFloat(s, 64); err == nil {
ts := int64(t * float64(time.Second))
return clientmodel.TimestampFromUnixNano(ts), nil
}
if t, err := time.Parse(time.RFC3339Nano, s); err == nil {
return clientmodel.TimestampFromTime(t), nil
}
return 0, fmt.Errorf("cannot parse %q to a valid timestamp", s)
}
func parseTimestampOrNow(t string, now clientmodel.Timestamp) (clientmodel.Timestamp, error) {
if t == "" {
return now, nil
}
tFloat, err := strconv.ParseFloat(t, 64)
if err != nil {
return 0, err
}
return clientmodel.TimestampFromUnixNano(int64(tFloat * float64(time.Second/time.Nanosecond))), nil
}
// QueryRange handles the /api/query_range endpoint.
func (serv MetricsService) QueryRange(w http.ResponseWriter, r *http.Request) {
setAccessControlHeaders(w)
w.Header().Set("Content-Type", "application/json")
params := httputils.GetQueryParams(r)
expr := params.Get("expr")
// Input times and durations are in seconds and get converted to nanoseconds.
endFloat, _ := strconv.ParseFloat(params.Get("end"), 64)
durationFloat, _ := strconv.ParseFloat(params.Get("range"), 64)
stepFloat, _ := strconv.ParseFloat(params.Get("step"), 64)
nanosPerSecond := int64(time.Second / time.Nanosecond)
end := int64(endFloat) * nanosPerSecond
duration := int64(durationFloat) * nanosPerSecond
step := int64(stepFloat) * nanosPerSecond
exprNode, err := rules.LoadExprFromString(expr)
if err != nil {
fmt.Fprint(w, ast.ErrorToJSON(err))
return
}
if exprNode.Type() != ast.VectorType {
fmt.Fprint(w, ast.ErrorToJSON(errors.New("expression does not evaluate to vector type")))
return
}
if end == 0 {
end = clientmodel.Now().UnixNano()
}
if step <= 0 {
step = nanosPerSecond
}
if end-duration < 0 {
duration = end
}
// For safety, limit the number of returned points per timeseries.
// This is sufficient for 60s resolution for a week or 1h resolution for a year.
if duration/step > 11000 {
fmt.Fprint(w, ast.ErrorToJSON(errors.New("exceeded maximum resolution of 11,000 points per timeseries. Try decreasing the query resolution (?step=XX)")))
return
}
// Align the start to step "tick" boundary.
end -= end % step
queryStats := stats.NewTimerGroup()
evalTimer := queryStats.GetTimer(stats.TotalEvalTime).Start()
matrix, err := ast.EvalVectorRange(
exprNode.(ast.VectorNode),
clientmodel.TimestampFromUnixNano(end-duration),
clientmodel.TimestampFromUnixNano(end),
time.Duration(step),
serv.Storage,
queryStats)
if err != nil {
fmt.Fprint(w, ast.ErrorToJSON(err))
return
}
evalTimer.Stop()
sortTimer := queryStats.GetTimer(stats.ResultSortTime).Start()
sort.Sort(matrix)
sortTimer.Stop()
jsonTimer := queryStats.GetTimer(stats.JSONEncodeTime).Start()
result := ast.TypedValueToJSON(matrix, "matrix")
jsonTimer.Stop()
glog.V(1).Infof("Range query: %s\nQuery stats:\n%s\n", expr, queryStats)
fmt.Fprint(w, result)
}
// NewTemplateExpander returns a template expander ready to use.
func NewTemplateExpander(text string, name string, data interface{}, timestamp clientmodel.Timestamp, queryEngine *promql.Engine, pathPrefix string) *templateExpander {
return &templateExpander{
text: text,
name: name,
data: data,
funcMap: text_template.FuncMap{
"query": func(q string) (queryResult, error) {
return query(q, timestamp, queryEngine)
},
"first": func(v queryResult) (*sample, error) {
if len(v) > 0 {
return v[0], nil
}
return nil, errors.New("first() called on vector with no elements")
},
"label": func(label string, s *sample) string {
return s.Labels[label]
},
"value": func(s *sample) float64 {
return s.Value
},
"strvalue": func(s *sample) string {
return s.Labels["__value__"]
},
"args": func(args ...interface{}) map[string]interface{} {
result := make(map[string]interface{})
for i, a := range args {
result[fmt.Sprintf("arg%d", i)] = a
}
return result
},
"reReplaceAll": func(pattern, repl, text string) string {
re := regexp.MustCompile(pattern)
return re.ReplaceAllString(text, repl)
},
"safeHtml": func(text string) html_template.HTML {
return html_template.HTML(text)
},
"match": regexp.MatchString,
"title": strings.Title,
"graphLink": strutil.GraphLinkForExpression,
"tableLink": strutil.TableLinkForExpression,
"sortByLabel": func(label string, v queryResult) queryResult {
sorter := queryResultByLabelSorter{v[:], label}
sort.Stable(sorter)
return v
},
"humanize": func(v float64) string {
if v == 0 || math.IsNaN(v) || math.IsInf(v, 0) {
return fmt.Sprintf("%.4g", v)
}
if math.Abs(v) >= 1 {
prefix := ""
for _, p := range []string{"k", "M", "G", "T", "P", "E", "Z", "Y"} {
if math.Abs(v) < 1000 {
break
}
prefix = p
v /= 1000
}
return fmt.Sprintf("%.4g%s", v, prefix)
}
prefix := ""
for _, p := range []string{"m", "u", "n", "p", "f", "a", "z", "y"} {
if math.Abs(v) >= 1 {
break
}
prefix = p
v *= 1000
}
return fmt.Sprintf("%.4g%s", v, prefix)
},
"humanize1024": func(v float64) string {
if math.Abs(v) <= 1 || math.IsNaN(v) || math.IsInf(v, 0) {
return fmt.Sprintf("%.4g", v)
}
prefix := ""
for _, p := range []string{"ki", "Mi", "Gi", "Ti", "Pi", "Ei", "Zi", "Yi"} {
if math.Abs(v) < 1024 {
break
}
prefix = p
v /= 1024
}
return fmt.Sprintf("%.4g%s", v, prefix)
},
"humanizeDuration": func(v float64) string {
if math.IsNaN(v) || math.IsInf(v, 0) {
return fmt.Sprintf("%.4g", v)
}
if v == 0 {
return fmt.Sprintf("%.4gs", v)
}
if math.Abs(v) >= 1 {
sign := ""
if v < 0 {
sign = "-"
v = -v
}
seconds := int64(v) % 60
minutes := (int64(v) / 60) % 60
hours := (int64(v) / 60 / 60) % 24
days := (int64(v) / 60 / 60 / 24)
// For days to minutes, we display seconds as an integer.
if days != 0 {
return fmt.Sprintf("%s%dd %dh %dm %ds", sign, days, hours, minutes, seconds)
}
if hours != 0 {
return fmt.Sprintf("%s%dh %dm %ds", sign, hours, minutes, seconds)
}
if minutes != 0 {
return fmt.Sprintf("%s%dm %ds", sign, minutes, seconds)
}
// For seconds, we display 4 significant digts.
return fmt.Sprintf("%s%.4gs", sign, v)
}
prefix := ""
for _, p := range []string{"m", "u", "n", "p", "f", "a", "z", "y"} {
if math.Abs(v) >= 1 {
break
}
prefix = p
v *= 1000
}
return fmt.Sprintf("%.4g%ss", v, prefix)
},
"humanizeTimestamp": func(v float64) string {
if math.IsNaN(v) || math.IsInf(v, 0) {
return fmt.Sprintf("%.4g", v)
}
t := clientmodel.TimestampFromUnixNano(int64(v * 1e9)).Time().UTC()
return fmt.Sprint(t)
},
"pathPrefix": func() string {
return pathPrefix
},
},
}
}
func extractHistogram(out Ingester, o *ProcessOptions, f *dto.MetricFamily) error {
samples := make(model.Samples, 0, len(f.Metric))
for _, m := range f.Metric {
if m.Histogram == nil {
continue
}
timestamp := o.Timestamp
if m.TimestampMs != nil {
timestamp = model.TimestampFromUnixNano(*m.TimestampMs * 1000000)
}
infSeen := false
for _, q := range m.Histogram.Bucket {
sample := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(q.GetCumulativeCount()),
Timestamp: timestamp,
}
samples = append(samples, sample)
metric := sample.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.LabelName(model.BucketLabel)] = model.LabelValue(fmt.Sprint(q.GetUpperBound()))
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_bucket")
if math.IsInf(q.GetUpperBound(), +1) {
infSeen = true
}
}
if m.Histogram.SampleSum != nil {
sum := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(m.Histogram.GetSampleSum()),
Timestamp: timestamp,
}
samples = append(samples, sum)
metric := sum.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_sum")
}
if m.Histogram.SampleCount != nil {
count := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(m.Histogram.GetSampleCount()),
Timestamp: timestamp,
}
samples = append(samples, count)
metric := count.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_count")
if !infSeen {
infBucket := &model.Sample{
Metric: model.Metric{},
Value: count.Value,
Timestamp: timestamp,
}
samples = append(samples, infBucket)
metric := infBucket.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.LabelName(model.BucketLabel)] = model.LabelValue("+Inf")
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_bucket")
}
}
}
return out.Ingest(samples)
}
func extractSummary(out Ingester, o *ProcessOptions, f *dto.MetricFamily) error {
samples := make(model.Samples, 0, len(f.Metric))
for _, m := range f.Metric {
if m.Summary == nil {
continue
}
timestamp := o.Timestamp
if m.TimestampMs != nil {
timestamp = model.TimestampFromUnixNano(*m.TimestampMs * 1000000)
}
for _, q := range m.Summary.Quantile {
sample := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(q.GetValue()),
Timestamp: timestamp,
}
samples = append(samples, sample)
metric := sample.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
// BUG(matt): Update other names to "quantile".
metric[model.LabelName(model.QuantileLabel)] = model.LabelValue(fmt.Sprint(q.GetQuantile()))
metric[model.MetricNameLabel] = model.LabelValue(f.GetName())
}
if m.Summary.SampleSum != nil {
sum := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(m.Summary.GetSampleSum()),
Timestamp: timestamp,
}
samples = append(samples, sum)
metric := sum.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_sum")
}
if m.Summary.SampleCount != nil {
count := &model.Sample{
Metric: model.Metric{},
Value: model.SampleValue(m.Summary.GetSampleCount()),
Timestamp: timestamp,
}
samples = append(samples, count)
metric := count.Metric
for _, p := range m.Label {
metric[model.LabelName(p.GetName())] = model.LabelValue(p.GetValue())
}
metric[model.MetricNameLabel] = model.LabelValue(f.GetName() + "_count")
}
}
return out.Ingest(samples)
}
