metrics: remove RR type (#4534)

To combat label cardinality explosions remove the type from metrics.
This was most severe in the histogram for request duration, remove it
there.

It's also highlighted difference between grpc and forward code, where
forward did use type and grpc didn't; getting rid of all that "fixes"
that discrepancy

Move monitor.go back into the vars directory and make it private again.
Also name it slightly better

Fixes: #4507

Signed-off-by: Miek Gieben <miek@miek.nl>

5 files changed, 13 insertions, 19 deletions

diff --git a/plugin/forward/connect.go b/plugin/forward/connect.go
index 7e34837e1..63a0bfe50 100644
--- a/plugin/forward/connect.go
+++ b/plugin/forward/connect.go
@@ -11,7 +11,6 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/coredns/coredns/plugin/pkg/dnsutil"
 	"github.com/coredns/coredns/request"
 
 	"github.com/miekg/dns"
@@ -130,11 +129,9 @@ func (p *Proxy) Connect(ctx context.Context, state request.Request, opts options
 		rc = strconv.Itoa(ret.Rcode)
 	}
 
-	qtype := dnsutil.QTypeMonitorLabel(state.QType())
-
 	RequestCount.WithLabelValues(p.addr).Add(1)
 	RcodeCount.WithLabelValues(rc, p.addr).Add(1)
-	RequestDuration.WithLabelValues(p.addr, rc, qtype).Observe(time.Since(start).Seconds())
+	RequestDuration.WithLabelValues(p.addr, rc).Observe(time.Since(start).Seconds())
 
 	return ret, nil
 }
diff --git a/plugin/forward/metrics.go b/plugin/forward/metrics.go
index 51685a599..9519c3c9b 100644
--- a/plugin/forward/metrics.go
+++ b/plugin/forward/metrics.go
@@ -27,7 +27,7 @@ var (
 		Name:      "request_duration_seconds",
 		Buckets:   plugin.TimeBuckets,
 		Help:      "Histogram of the time each request took.",
-	}, []string{"to", "rcode", "type"})
+	}, []string{"to", "rcode"})
 	HealthcheckFailureCount = promauto.NewCounterVec(prometheus.CounterOpts{
 		Namespace: plugin.Namespace,
 		Subsystem: "forward",
diff --git a/plugin/pkg/dnsutil/monitor.go b/plugin/metrics/vars/monitor.go
index b86f8e43a..17ca7cf63 100644
--- a/plugin/pkg/dnsutil/monitor.go
+++ b/plugin/metrics/vars/monitor.go
@@ -1,4 +1,4 @@
-package dnsutil
+package vars
 
 import (
 	"github.com/miekg/dns"
@@ -25,13 +25,11 @@ var monitorType = map[uint16]struct{}{
 	dns.TypeANY:  {},
 }
 
-const other = "other"
-
-// QTypeMonitorLabel returns dns type label based on a list of monitored types.
-// Will return "other" for unmonitored ones.
-func QTypeMonitorLabel(qtype uint16) string {
+// qTypeString returns the RR type based on monitorType. It returns the text representation
+// of thosAe types. RR types not in that list will have "other" returned.
+func qTypeString(qtype uint16) string {
 	if _, known := monitorType[qtype]; known {
 		return dns.Type(qtype).String()
 	}
-	return other
+	return "other"
 }
diff --git a/plugin/metrics/vars/report.go b/plugin/metrics/vars/report.go
index 9bad43582..d32e361b1 100644
--- a/plugin/metrics/vars/report.go
+++ b/plugin/metrics/vars/report.go
@@ -3,7 +3,6 @@ package vars
 import (
 	"time"
 
-	"github.com/coredns/coredns/plugin/pkg/dnsutil"
 	"github.com/coredns/coredns/request"
 )
 
@@ -18,14 +17,14 @@ func Report(server string, req request.Request, zone, rcode string, size int, st
 		fam = "2"
 	}
 
-	qtype := dnsutil.QTypeMonitorLabel(req.QType())
-
 	if req.Do() {
 		RequestDo.WithLabelValues(server, zone).Inc()
 	}
 
+	qtype := qTypeString(req.QType())
 	RequestCount.WithLabelValues(server, zone, net, fam, qtype).Inc()
-	RequestDuration.WithLabelValues(server, zone, qtype).Observe(time.Since(start).Seconds())
+
+	RequestDuration.WithLabelValues(server, zone).Observe(time.Since(start).Seconds())
 
 	ResponseSize.WithLabelValues(server, zone, net).Observe(float64(size))
 	RequestSize.WithLabelValues(server, zone, net).Observe(float64(req.Len()))
diff --git a/plugin/metrics/vars/vars.go b/plugin/metrics/vars/vars.go
index 2187e13f0..7a803a62a 100644
--- a/plugin/metrics/vars/vars.go
+++ b/plugin/metrics/vars/vars.go
@@ -21,14 +21,14 @@ var (
 		Subsystem: subsystem,
 		Name:      "request_duration_seconds",
 		Buckets:   plugin.TimeBuckets,
-		Help:      "Histogram of the time (in seconds) each request took.",
-	}, []string{"server", "zone", "type"})
+		Help:      "Histogram of the time (in seconds) each request took per zone.",
+	}, []string{"server", "zone"})
 
 	RequestSize = promauto.NewHistogramVec(prometheus.HistogramOpts{
 		Namespace: plugin.Namespace,
 		Subsystem: subsystem,
 		Name:      "request_size_bytes",
-		Help:      "Size of the EDNS0 UDP buffer in bytes (64K for TCP).",
+		Help:      "Size of the EDNS0 UDP buffer in bytes (64K for TCP) per zone and protocol.",
 		Buckets:   []float64{0, 100, 200, 300, 400, 511, 1023, 2047, 4095, 8291, 16e3, 32e3, 48e3, 64e3},
 	}, []string{"server", "zone", "proto"})