From 6c639be8ecf91885326eb7a51ee6689954c42211 Mon Sep 17 00:00:00 2001
From: Matt Keeler <mkeeler@users.noreply.github.com>
Date: Fri, 23 Apr 2021 17:05:33 -0400
Subject: [PATCH] Add prometheus guage definitions for replication metrics.
 (#10109)

---
 agent/consul/replication.go | 52 +++++++++++++++++++++++++++++++++++++
 agent/setup.go              |  1 +
 2 files changed, 53 insertions(+)

diff --git a/agent/consul/replication.go b/agent/consul/replication.go
index 9ad3065cf..7a861901d 100644
--- a/agent/consul/replication.go
+++ b/agent/consul/replication.go
@@ -7,6 +7,7 @@ import (
 	"time"
 
 	metrics "github.com/armon/go-metrics"
+	"github.com/armon/go-metrics/prometheus"
 	"github.com/hashicorp/go-hclog"
 	"golang.org/x/time/rate"
 
@@ -14,6 +15,57 @@ import (
 	"github.com/hashicorp/consul/logging"
 )
 
+var ReplicationGauges = []prometheus.GaugeDefinition{
+	{
+		Name: []string{"leader", "replication", "acl-policies", "status"},
+		Help: "Tracks the current health of ACL policy replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "acl-policies", "index"},
+		Help: "Tracks the index of ACL policies in the primary that the secondary has successfully replicated",
+	},
+	{
+		Name: []string{"leader", "replication", "acl-tokens", "status"},
+		Help: "Tracks the current health of ACL token replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "acl-tokens", "index"},
+		Help: "Tracks the index of ACL tokens in the primary that the secondary has successfully replicated",
+	},
+	{
+		Name: []string{"leader", "replication", "acl-roles", "status"},
+		Help: "Tracks the current health of ACL role replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "acl-roles", "index"},
+		Help: "Tracks the index of ACL roles in the primary that the secondary has successfully replicated",
+	},
+	{
+		Name: []string{"leader", "replication", "config-entries", "status"},
+		Help: "Tracks the current health of config entry replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "config-entries", "index"},
+		Help: "Tracks the index of config entries in the primary that the secondary has successfully replicated",
+	},
+	{
+		Name: []string{"leader", "replication", "federation-state", "status"},
+		Help: "Tracks the current health of federation state replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "federation-state", "index"},
+		Help: "Tracks the index of federation states in the primary that the secondary has successfully replicated",
+	},
+	{
+		Name: []string{"leader", "replication", "namespaces", "status"},
+		Help: "Tracks the current health of federation state replication on the leader",
+	},
+	{
+		Name: []string{"leader", "replication", "namespaces", "index"},
+		Help: "Tracks the index of federation states in the primary that the secondary has successfully replicated",
+	},
+}
+
 const (
 	// replicationMaxRetryWait is the maximum number of seconds to wait between
 	// failed blocking queries when backing off.
diff --git a/agent/setup.go b/agent/setup.go
index 0d485f08a..9a84e7d24 100644
--- a/agent/setup.go
+++ b/agent/setup.go
@@ -202,6 +202,7 @@ func getPrometheusDefs(cfg lib.TelemetryConfig) ([]prometheus.GaugeDefinition, [
 		consul.SessionGauges,
 		grpc.StatsGauges,
 		usagemetrics.Gauges,
+		consul.ReplicationGauges,
 		Gauges,
 	}
 	// Flatten definitions