Extend metric vm_promscrape_targets with status label (#615)

The change to `vm_promscrape_targets` metric suppose to improve observability
for `vmagent` so it will be possible to track how many targets are up or down
for every specific scrape group:
```
vm_promscrape_targets{type="static_configs", status="down"} 1
vm_promscrape_targets{type="static_configs", status="up"} 2
```

lib/promscrape/scraper.go

@@ -236,11 +236,11 @@ func newScraperGroup(name string, pushData func(wr *prompbmarshal.WriteRequest))
 		pushData:     pushData,
 		changesCount: metrics.NewCounter(fmt.Sprintf(`vm_promscrape_config_changes_total{type=%q}`, name)),
 	}
-	metrics.NewGauge(fmt.Sprintf(`vm_promscrape_targets{type=%q}`, name), func() float64 {
-		sg.mLock.Lock()
-		n := len(sg.m)
-		sg.mLock.Unlock()
-		return float64(n)
+	metrics.NewGauge(fmt.Sprintf(`vm_promscrape_targets{type=%q, status="up"}`, name), func() float64 {
+		return float64(tsmGlobal.StatusByGroup(sg.name, true))
+	})
+	metrics.NewGauge(fmt.Sprintf(`vm_promscrape_targets{type=%q, status="down"}`, name), func() float64 {
+		return float64(tsmGlobal.StatusByGroup(sg.name, false))
 	})
 	return sg
 }
@@ -277,7 +277,7 @@ func (sg *scraperGroup) update(sws []ScrapeWork) {
 		}
 
 		// Start a scraper for the missing key.
-		sc := newScraper(sw, sg.pushData)
+		sc := newScraper(sw, sg.name, sg.pushData)
 		sg.wg.Add(1)
 		go func() {
 			defer sg.wg.Done()
@@ -309,12 +309,13 @@ type scraper struct {
 	stopCh chan struct{}
 }
 
-func newScraper(sw *ScrapeWork, pushData func(wr *prompbmarshal.WriteRequest)) *scraper {
+func newScraper(sw *ScrapeWork, group string, pushData func(wr *prompbmarshal.WriteRequest)) *scraper {
 	sc := &scraper{
 		stopCh: make(chan struct{}),
 	}
 	c := newClient(sw)
 	sc.sw.Config = *sw
+	sc.sw.ScrapeGroup = group
 	sc.sw.ReadData = c.ReadData
 	sc.sw.PushData = pushData
 	return sc

lib/promscrape/scrapework.go

@@ -120,6 +120,10 @@ type scrapeWork struct {
 	// PushData is called for pushing collected data.
 	PushData func(wr *prompbmarshal.WriteRequest)
 
+	// ScrapeGroup is name of ScrapeGroup that
+	// scrapeWork belongs to
+	ScrapeGroup string
+
 	bodyBuf []byte
 	rows    parser.Rows
 	tmpRow  parser.Row
@@ -232,7 +236,7 @@ func (sw *scrapeWork) scrapeInternal(timestamp int64) error {
 	prompbmarshal.ResetWriteRequest(&sw.writeRequest)
 	sw.labels = sw.labels[:0]
 	sw.samples = sw.samples[:0]
-	tsmGlobal.Update(&sw.Config, up == 1, timestamp, int64(duration*1000), err)
+	tsmGlobal.Update(&sw.Config, sw.ScrapeGroup, up == 1, timestamp, int64(duration*1000), err)
 	return err
 }
 

lib/promscrape/targetstatus.go

@@ -46,11 +46,12 @@ func (tsm *targetStatusMap) Unregister(sw *ScrapeWork) {
 	tsm.mu.Unlock()
 }
 
-func (tsm *targetStatusMap) Update(sw *ScrapeWork, up bool, scrapeTime, scrapeDuration int64, err error) {
+func (tsm *targetStatusMap) Update(sw *ScrapeWork, group string, up bool, scrapeTime, scrapeDuration int64, err error) {
 	tsm.mu.Lock()
 	tsm.m[sw.ID] = targetStatus{
 		sw:             sw,
 		up:             up,
+		scrapeGroup:    group,
 		scrapeTime:     scrapeTime,
 		scrapeDuration: scrapeDuration,
 		err:            err,
@@ -58,6 +59,20 @@ func (tsm *targetStatusMap) Update(sw *ScrapeWork, up bool, scrapeTime, scrapeDu
 	tsm.mu.Unlock()
 }
 
+// StatusByGroup returns the number of targets with status==up
+// for the given group name
+func (tsm *targetStatusMap) StatusByGroup(group string, up bool) int {
+	var count int
+	tsm.mu.Lock()
+	for _, st := range tsm.m {
+		if st.scrapeGroup == group && st.up == up {
+			count++
+		}
+	}
+	tsm.mu.Unlock()
+	return count
+}
+
 func (tsm *targetStatusMap) WriteHumanReadable(w io.Writer) {
 	byJob := make(map[string][]targetStatus)
 	tsm.mu.Lock()
@@ -116,6 +131,7 @@ type jobStatus struct {
 type targetStatus struct {
 	sw             *ScrapeWork
 	up             bool
+	scrapeGroup    string
 	scrapeTime     int64
 	scrapeDuration int64
 	err            error