From 4d071bed24c185daa0da1ee319256019fff75e3f Mon Sep 17 00:00:00 2001
From: raul-te <46945030+raul-te@users.noreply.github.com>
Date: Fri, 22 May 2020 00:11:00 -0700
Subject: [PATCH] Add option to disable mongodb cluster status (#7515)

It can be expensive to compute these metrics. In particular, when
retrieveing the amount of jumbo chunks, an index is not being used and
consequently the query triggers an expensive COLLSCAN. For big
databases, this query has negative impact on the cluster performance.
---
 plugins/inputs/mongodb/README.md         |  5 +++++
 plugins/inputs/mongodb/mongodb.go        | 27 ++++++++++++++++--------
 plugins/inputs/mongodb/mongodb_server.go | 12 +++++++----
 3 files changed, 31 insertions(+), 13 deletions(-)

diff --git a/plugins/inputs/mongodb/README.md b/plugins/inputs/mongodb/README.md
index 5e0f304d6..cce93dc07 100644
--- a/plugins/inputs/mongodb/README.md
+++ b/plugins/inputs/mongodb/README.md
@@ -11,6 +11,11 @@
   ##   mongodb://10.10.3.33:18832,
   servers = ["mongodb://127.0.0.1:27017"]
 
+  ## When true, collect cluster status.
+  ## Note that the query that counts jumbo chunks triggers a COLLSCAN, which
+  ## may have an impact on performance.
+  # gather_cluster_status = true
+
   ## When true, collect per database stats
   # gather_perdb_stats = false
 
diff --git a/plugins/inputs/mongodb/mongodb.go b/plugins/inputs/mongodb/mongodb.go
index 967ccbe5f..016515ea9 100644
--- a/plugins/inputs/mongodb/mongodb.go
+++ b/plugins/inputs/mongodb/mongodb.go
@@ -17,12 +17,13 @@ import (
 )
 
 type MongoDB struct {
-	Servers          []string
-	Ssl              Ssl
-	mongos           map[string]*Server
-	GatherPerdbStats bool
-	GatherColStats   bool
-	ColStatsDbs      []string
+	Servers             []string
+	Ssl                 Ssl
+	mongos              map[string]*Server
+	GatherClusterStatus bool
+	GatherPerdbStats    bool
+	GatherColStats      bool
+	ColStatsDbs         []string
 	tlsint.ClientConfig
 
 	Log telegraf.Logger
@@ -41,6 +42,11 @@ var sampleConfig = `
   ##   mongodb://10.10.3.33:18832,
   servers = ["mongodb://127.0.0.1:27017"]
 
+  ## When true, collect cluster status
+  ## Note that the query that counts jumbo chunks triggers a COLLSCAN, which
+  ## may have an impact on performance.
+  # gather_cluster_status = true
+
   ## When true, collect per database stats
   # gather_perdb_stats = false
 
@@ -177,14 +183,17 @@ func (m *MongoDB) gatherServer(server *Server, acc telegraf.Accumulator) error {
 		}
 		server.Session = sess
 	}
-	return server.gatherData(acc, m.GatherPerdbStats, m.GatherColStats, m.ColStatsDbs)
+	return server.gatherData(acc, m.GatherClusterStatus, m.GatherPerdbStats, m.GatherColStats, m.ColStatsDbs)
 }
 
 func init() {
 	inputs.Add("mongodb", func() telegraf.Input {
 		return &MongoDB{
-			ColStatsDbs: []string{"local"},
-			mongos:      make(map[string]*Server),
+			mongos:              make(map[string]*Server),
+			GatherClusterStatus: true,
+			GatherPerdbStats:    false,
+			GatherColStats:      false,
+			ColStatsDbs:         []string{"local"},
 		}
 	})
 }
diff --git a/plugins/inputs/mongodb/mongodb_server.go b/plugins/inputs/mongodb/mongodb_server.go
index be3916b5e..5af48c10a 100644
--- a/plugins/inputs/mongodb/mongodb_server.go
+++ b/plugins/inputs/mongodb/mongodb_server.go
@@ -192,7 +192,7 @@ func (s *Server) gatherCollectionStats(colStatsDbs []string) (*ColStats, error)
 	return results, nil
 }
 
-func (s *Server) gatherData(acc telegraf.Accumulator, gatherDbStats bool, gatherColStats bool, colStatsDbs []string) error {
+func (s *Server) gatherData(acc telegraf.Accumulator, gatherClusterStatus bool, gatherDbStats bool, gatherColStats bool, colStatsDbs []string) error {
 	s.Session.SetMode(mgo.Eventual, true)
 	s.Session.SetSocketTimeout(0)
 
@@ -218,9 +218,13 @@ func (s *Server) gatherData(acc telegraf.Accumulator, gatherDbStats bool, gather
 		}
 	}
 
-	clusterStatus, err := s.gatherClusterStatus()
-	if err != nil {
-		s.Log.Debugf("Unable to gather cluster status: %s", err.Error())
+	var clusterStatus *ClusterStatus
+	if gatherClusterStatus {
+		status, err := s.gatherClusterStatus()
+		if err != nil {
+			s.Log.Debugf("Unable to gather cluster status: %s", err.Error())
+		}
+		clusterStatus = status
 	}
 
 	shardStats, err := s.gatherShardConnPoolStats()