app/vmagent: reduce memory usage when importing data via /api/v1/import

Previously vmagent could use big amounts of RAM when each ingested JSON line
contained many samples.

app/vmagent/remotewrite/pendingseries.go

@@ -19,7 +19,7 @@ var (
 	flushInterval = flag.Duration("remoteWrite.flushInterval", time.Second, "Interval for flushing the data to remote storage. "+
 		"Higher value reduces network bandwidth usage at the cost of delayed push of scraped data to remote storage. "+
 		"Minimum supported interval is 1 second")
-	maxUnpackedBlockSize = flagutil.NewBytes("remoteWrite.maxBlockSize", 32*1024*1024, "The maximum size in bytes of unpacked request to send to remote storage. "+
+	maxUnpackedBlockSize = flagutil.NewBytes("remoteWrite.maxBlockSize", 8*1024*1024, "The maximum size in bytes of unpacked request to send to remote storage. "+
 		"It shouldn't exceed -maxInsertRequestSize from VictoriaMetrics")
 )
 
@@ -127,7 +127,7 @@ func (wr *writeRequest) push(src []prompbmarshal.TimeSeries) {
 	for i := range src {
 		tssDst = append(tssDst, prompbmarshal.TimeSeries{})
 		wr.copyTimeSeries(&tssDst[len(tssDst)-1], &src[i])
-		if len(tssDst) >= maxRowsPerBlock {
+		if len(wr.samples) >= maxRowsPerBlock {
 			wr.tss = tssDst
 			wr.flush()
 			tssDst = wr.tss

app/vmagent/remotewrite/remotewrite.go

@@ -153,10 +153,19 @@ func Push(wr *prompbmarshal.WriteRequest) {
 	tss := wr.Timeseries
 	for len(tss) > 0 {
 		// Process big tss in smaller blocks in order to reduce the maximum memory usage
+		samplesCount := 0
+		i := 0
+		for i < len(tss) {
+			samplesCount += len(tss[i].Samples)
+			i++
+			if samplesCount > maxRowsPerBlock {
+				break
+			}
+		}
 		tssBlock := tss
-		if len(tssBlock) > maxRowsPerBlock {
+		if i < len(tss) {
-			tssBlock = tss[:maxRowsPerBlock]
+			tssBlock = tss[:i]
-			tss = tss[maxRowsPerBlock:]
+			tss = tss[i:]
 		} else {
 			tss = nil
 		}