From b371eb679dcf9dd14a5a819f80b3e3b624fdfe7f Mon Sep 17 00:00:00 2001
From: Miek Gieben <miek@miek.nl>
Date: Sat, 18 Jan 2020 08:39:02 +0100
Subject: [PATCH] Add metrics

Signed-off-by: Miek Gieben <miek@miek.nl>
---
 plugin/traffic/README.md      |  8 +++++++-
 plugin/traffic/setup.go       |  2 ++
 plugin/traffic/xds/client.go  |  1 +
 plugin/traffic/xds/metrics.go | 17 +++++++++++++++++
 4 files changed, 27 insertions(+), 1 deletion(-)
 create mode 100644 plugin/traffic/xds/metrics.go

diff --git a/plugin/traffic/README.md b/plugin/traffic/README.md
index e973a7968..968d43d2d 100644
--- a/plugin/traffic/README.md
+++ b/plugin/traffic/README.md
@@ -91,6 +91,13 @@ What metrics should we do? If any? Number of clusters? Number of endpoints and h
 This plugin report readiness to the ready plugin. This will happen after a gRPC stream has been
 established to the control plane.
 
+## Metrics
+
+If monitoring is enabled (via the *prometheus* plugin) then the following metric are exported:
+
+* `coredns_traffic_clusters_tracked{}` the number of tracked clusters.
+
+
 ## Examples
 
 ~~~
@@ -123,7 +130,6 @@ is not implemented.
 
 ## TODO
 
-* metrics?
 * credentials (other than TLS) - how/what?
 * is the protocol correctly implemented? Should we not have a 10s tick, but wait for responses from
   the control plane?
diff --git a/plugin/traffic/setup.go b/plugin/traffic/setup.go
index bfd9683a4..4557faa7e 100644
--- a/plugin/traffic/setup.go
+++ b/plugin/traffic/setup.go
@@ -9,6 +9,7 @@ import (
 
 	"github.com/coredns/coredns/core/dnsserver"
 	"github.com/coredns/coredns/plugin"
+	"github.com/coredns/coredns/plugin/metrics"
 	clog "github.com/coredns/coredns/plugin/pkg/log"
 	"github.com/coredns/coredns/plugin/pkg/parse"
 	pkgtls "github.com/coredns/coredns/plugin/pkg/tls"
@@ -38,6 +39,7 @@ func setup(c *caddy.Controller) error {
 
 	c.OnStartup(func() error {
 		go t.c.Run()
+		metrics.MustRegister(c, xds.ClusterGauge)
 		return nil
 	})
 	c.OnShutdown(func() error { return t.c.Stop() })
diff --git a/plugin/traffic/xds/client.go b/plugin/traffic/xds/client.go
index d70832de7..acb395a12 100644
--- a/plugin/traffic/xds/client.go
+++ b/plugin/traffic/xds/client.go
@@ -191,6 +191,7 @@ func (c *Client) receive(stream adsStream) error {
 				a.SetClusterLoadAssignment(cluster.GetName(), nil)
 			}
 			log.Debugf("Cluster discovery processed with %d resources, version %q and nonce %q", len(resp.GetResources()), c.Version(cdsURL), c.Nonce(cdsURL))
+			ClusterGauge.WithLabelValues().Set(float64(len(resp.GetResources())))
 			// set our local administration and ack the reply. Empty version would signal NACK.
 			c.SetNonce(cdsURL, resp.GetNonce())
 			c.SetVersion(cdsURL, resp.GetVersionInfo())
diff --git a/plugin/traffic/xds/metrics.go b/plugin/traffic/xds/metrics.go
new file mode 100644
index 000000000..0d9175497
--- /dev/null
+++ b/plugin/traffic/xds/metrics.go
@@ -0,0 +1,17 @@
+package xds
+
+import (
+	"github.com/coredns/coredns/plugin"
+
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+var (
+	// ClusterGauge is the number of clusters we are currently tracking.
+	ClusterGauge = prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: plugin.Namespace,
+		Subsystem: "traffic",
+		Name:      "clusters_tracked",
+		Help:      "Gauge of tracked clusters.",
+	}, []string{""})
+)