Optionally enhance "sources" information with NtpData response (#106)

This commit extends the "sources" information with additional peer/source
information from the NtpData response. Namely it adds:

- Offset
- Peer Delay
- Peer Dispersion
- Peer Response Time
- Peer Jitter Asymmetry

The additional data collection is hidden behind a new flag
`--collector.sources-with-ntpdata` which defaults to `false`

Signed-off-by: Rudolph Bott <[email protected]>

Author: rbott

README.md

@@ -60,6 +60,8 @@ Flags:
       --chrony.timeout=5s        Timeout on requests to the Chrony srever.
       --[no-]collector.tracking  Collect tracking metrics
       --[no-]collector.sources   Collect sources metrics
+      --[no-]collector.sources.with-ntpdata  
+                                 Extend sources with ntpdata metrics (requires socket connection)
       --[no-]collector.serverstats  
                                  Collect serverstats metrics
       --[no-]collector.chmod-socket  

collector/collector.go

@@ -55,6 +55,7 @@ type Exporter struct {
 	timeout time.Duration
 
 	collectSources     bool
+	collectNtpdata     bool
 	collectTracking    bool
 	collectServerstats bool
 	chmodSocket        bool
@@ -86,6 +87,8 @@ type ChronyCollectorConfig struct {
 
 	// CollectSources will configure the exporter to collect `chronyc sources`.
 	CollectSources bool
+	// CollectNtpData will configure the exporter to extend sources info with `chronyc ntpdata`
+	CollectNtpdata bool
 	// CollectTracking will configure the exporter to collect `chronyc tracking`.
 	CollectTracking bool
 	// CollectServerstats will configure the exporter to collect `chronyc serverstats`.
@@ -98,6 +101,7 @@ func NewExporter(conf ChronyCollectorConfig, logger *slog.Logger) Exporter {
 		timeout: conf.Timeout,
 
 		collectSources:     conf.CollectSources,
+		collectNtpdata:     conf.CollectNtpdata,
 		collectTracking:    conf.CollectTracking,
 		collectServerstats: conf.CollectServerstats,
 		chmodSocket:        conf.ChmodSocket,
@@ -161,7 +165,7 @@ func (e Exporter) Collect(ch chan<- prometheus.Metric) {
 	client := chrony.Client{Sequence: 1, Connection: conn}
 
 	if e.collectSources {
-		err = e.getSourcesMetrics(logger, ch, client)
+		err = e.getSourcesMetrics(logger, ch, client, e.collectNtpdata)
 		if err != nil {
 			logger.Debug("Couldn't get sources", "err", err)
 			up = 0

collector/sources.go

@@ -108,9 +108,59 @@ var (
 		),
 		prometheus.GaugeValue,
 	}
+
+	sourcesPeerOffset = typedDesc{
+		prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, sourcesSubsystem, "peer_offset_seconds"),
+			"Chrony sources peer offset",
+			[]string{"source_address", "source_name"},
+			nil,
+		),
+		prometheus.GaugeValue,
+	}
+
+	sourcesPeerDelay = typedDesc{
+		prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, sourcesSubsystem, "peer_delay_seconds"),
+			"Chrony sources peer delay",
+			[]string{"source_address", "source_name"},
+			nil,
+		),
+		prometheus.GaugeValue,
+	}
+
+	sourcesPeerDispersion = typedDesc{
+		prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, sourcesSubsystem, "peer_dispersion_seconds"),
+			"Chrony sources peer dispersion",
+			[]string{"source_address", "source_name"},
+			nil,
+		),
+		prometheus.GaugeValue,
+	}
+
+	sourcesPeerResponseTime = typedDesc{
+		prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, sourcesSubsystem, "peer_response_time_seconds"),
+			"Chrony sources peer response time",
+			[]string{"source_address", "source_name"},
+			nil,
+		),
+		prometheus.GaugeValue,
+	}
+
+	sourcesPeerJitterAsymmetry = typedDesc{
+		prometheus.NewDesc(
+			prometheus.BuildFQName(namespace, sourcesSubsystem, "peer_jitter_asymmetry_seconds"),
+			"Chrony sources peer jitter asymmetry",
+			[]string{"source_address", "source_name"},
+			nil,
+		),
+		prometheus.GaugeValue,
+	}
 )
 
-func (e Exporter) getSourcesMetrics(logger *slog.Logger, ch chan<- prometheus.Metric, client chrony.Client) error {
+func (e Exporter) getSourcesMetrics(logger *slog.Logger, ch chan<- prometheus.Metric, client chrony.Client, collectNtpdata bool) error {
 	packet, err := client.Communicate(chrony.NewSourcesPacket())
 	if err != nil {
 		return err
@@ -157,6 +207,22 @@ func (e Exporter) getSourcesMetrics(logger *slog.Logger, ch chan<- prometheus.Me
 		ch <- sourcesPollInterval.mustNewConstMetric(math.Pow(2, float64(r.Poll)), sourceAddress, sourceName)
 		ch <- sourcesStateInfo.mustNewConstMetric(1.0, sourceAddress, sourceName, r.State.String(), r.Mode.String())
 		ch <- sourcesStratum.mustNewConstMetric(float64(r.Stratum), sourceAddress, sourceName)
+
+		if collectNtpdata {
+			ntpDataPacket, err := client.Communicate(chrony.NewNTPDataPacket(r.IPAddr))
+			if err != nil {
+				return fmt.Errorf("Failed to get ntpdata response for: %s", r.IPAddr)
+			}
+			ntpData, ok := ntpDataPacket.(*chrony.ReplyNTPData)
+			if !ok {
+				return fmt.Errorf("Got wrong 'ntpdata' response: %q", packet)
+			}
+			ch <- sourcesPeerOffset.mustNewConstMetric(ntpData.Offset, sourceAddress, sourceName)
+			ch <- sourcesPeerDelay.mustNewConstMetric(ntpData.PeerDelay, sourceAddress, sourceName)
+			ch <- sourcesPeerResponseTime.mustNewConstMetric(ntpData.ResponseTime, sourceAddress, sourceName)
+			ch <- sourcesPeerDispersion.mustNewConstMetric(ntpData.PeerDispersion, sourceAddress, sourceName)
+			ch <- sourcesPeerJitterAsymmetry.mustNewConstMetric(ntpData.JitterAsymmetry, sourceAddress, sourceName)
+		}
 	}
 
 	return nil

main.go

@@ -57,6 +57,11 @@ func main() {
 		"Collect sources metrics",
 	).Default("false").BoolVar(&conf.CollectSources)
 
+	kingpin.Flag(
+		"collector.sources.with-ntpdata",
+		"Extend sources with ntpdata metrics (requires socket connection)",
+	).Default("false").BoolVar(&conf.CollectNtpdata)
+
 	kingpin.Flag(
 		"collector.serverstats",
 		"Collect serverstats metrics",