middleware/proxy: drop protcol in metric

This blows up prometheus (panic in CoreDNS) and is not that
handy to have.

2 files changed, 4 insertions, 5 deletions

diff --git a/middleware/proxy/README.md b/middleware/proxy/README.md
index 2bd6c7143..9ca0e4638 100644
--- a/middleware/proxy/README.md
+++ b/middleware/proxy/README.md
@@ -60,11 +60,10 @@ Currently supported are `dns` (i.e., standard DNS over UDP) and `https_google`.
 
 If monitoring is enabled (via the *prometheus* directive) then the following metric is exported:
 
-* coredns_proxy_request_count_total{protocol, zone, family}
+* coredns_proxy_request_duration_milliseconds{zone}
 
-This has some overlap with `coredns_dns_request_count_total{zone, proto, family}`, but allows for
-specifics on upstream query resolving. See the *prometheus* documentation for more details.
-`protocol` is the protocol used to query the upstream.
+The metric shows the duration for a proxied request, the `zone` label is the **FROM** as specified
+in the configuration.
 
 ## Examples
 
diff --git a/middleware/proxy/metrics.go b/middleware/proxy/metrics.go
index a2205c692..77d268bd1 100644
--- a/middleware/proxy/metrics.go
+++ b/middleware/proxy/metrics.go
@@ -16,7 +16,7 @@ var (
 		Name:      "request_duration_milliseconds",
 		Buckets:   append(prometheus.DefBuckets, []float64{50, 100, 200, 500, 1000, 2000, 3000, 4000, 5000, 10000}...),
 		Help:      "Histogram of the time (in milliseconds) each request took.",
-	}, []string{"protocol", "zone"})
+	}, []string{"zone"})
 )
 
 // OnStartup sets up the metrics on startup. This is done for all proxy protocols.