app/vmagent/remotewrite: expose vmagent_remotewrite_send_duration_seconds_total metric

This metric can be used for determining high saturation of every connection to remote storage with
an alerting query `rate(vmagent_remotewrite_send_duration_seconds_total) > 0.9s`.
This query triggers when a connection is satureated by more than 90%

app/vmagent/remotewrite/client.go

@@ -78,6 +78,7 @@ type client struct {
 	errorsCount     *metrics.Counter
 	packetsDropped  *metrics.Counter
 	retriesCount    *metrics.Counter
+	sendDuration    *metrics.FloatCounter
 
 	wg     sync.WaitGroup
 	stopCh chan struct{}
@@ -133,6 +134,7 @@ func newClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persistentqu
 	c.errorsCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_errors_total{url=%q}`, c.sanitizedURL))
 	c.packetsDropped = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_packets_dropped_total{url=%q}`, c.sanitizedURL))
 	c.retriesCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_retries_count_total{url=%q}`, c.sanitizedURL))
+	c.sendDuration = metrics.GetOrCreateFloatCounter(fmt.Sprintf(`vmagent_remotewrite_send_duration_seconds_total{url=%q}`, c.sanitizedURL))
 	for i := 0; i < concurrency; i++ {
 		c.wg.Add(1)
 		go func() {
@@ -204,7 +206,9 @@ func (c *client) runWorker() {
 			return
 		}
 		go func() {
+			startTime := time.Now()
 			ch <- c.sendBlock(block)
+			c.sendDuration.Add(time.Since(startTime).Seconds())
 		}()
 		select {
 		case ok := <-ch:

docs/CHANGELOG.md

@@ -16,7 +16,8 @@ sort: 15
 * FEATURE: add `-search.maxSamplesPerQuery` command-line flag for limiting the number of raw samples a single query can process across all the time series. This option can protect from heavy queries, which select too big number of raw samples. Thanks to @jiangxinlingdu for [the initial pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/1478).
 * FEATURE: improve performance for queries that process big number of time series and/or samples on systems with big number of CPU cores.
 * FEATURE: vmalert: expose `vmalert_alerting_rules_last_evaluation_samples` and `vmalert_recording_rules_last_evaluation_samples` metrics. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1494).
-* FEATURE: vminsert: expose `vm_rpc_send_duration_seconds_total` counter, which can be used for determining high saturation of every `vminsert -> vmstorage` link with an alerting query `rate(vm_rpc_send_duration_seconds_total) > 0.9s`. This query triggers when the link is saturated by more than 90%.
+* FEATURE: vminsert: expose `vm_rpc_send_duration_seconds_total` counter, which can be used for determining high saturation of every `vminsert -> vmstorage` link with an alerting query `rate(vm_rpc_send_duration_seconds_total) > 0.9s`. This query triggers when the link is saturated by more than 90%. This usually means that more `vminsert` or `vmstorage` nodes must be added to the cluster in order to increase the total number of `vminsert -> vmstorage` links.
+* FEATURE: vmagent: expose `vmagent_remotewrite_send_duration_seconds_total` counter, which can be used for determining high saturation of every connection to remote storage with an alerting query `rate(vmagent_remotewrite_send_duration_seconds_total) > 0.9s`. This query triggers when a connection is satureated by more than 90%. This usually means that `-remoteWrite.queues` command-line flag must be increased in order to increase the number of connections per each remote storage.
 
 * BUGFIX: fix corner cases for queries on time ranges exceeding 40 days. Previously some series can be missing in query results. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1486).
 * BUGFIX: vmselect: return dummy response at `/rules` page in the same way as for `/api/v1/rules` page. The `/rules` page is requested by Grafana 8. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1493) for details.