Added Consul health checks state monitoring. (#1294)

CHANGELOG.md

@@ -22,7 +22,7 @@ in conjunction with wildcard dimension values as it will control the amount of
 time before a new metric is included by the plugin.
 
 ### Features
-
+- [#1294](https://github.com/influxdata/telegraf/pull/1294): consul input plugin. Thanks @harnash
 - [#1164](https://github.com/influxdata/telegraf/pull/1164): conntrack input plugin. Thanks @robinpercy!
 - [#1165](https://github.com/influxdata/telegraf/pull/1165): vmstat input plugin. Thanks @jshim-xm!
 - [#1247](https://github.com/influxdata/telegraf/pull/1247): rollbar input plugin. Thanks @francois2metz and @cduez!

Godeps

@@ -23,6 +23,7 @@ github.com/gonuts/go-shellquote e842a11b24c6abfb3dd27af69a17f482e4b483c2
 github.com/gorilla/context 1ea25387ff6f684839d82767c1733ff4d4d15d0a
 github.com/gorilla/mux c9e326e2bdec29039a3761c07bece13133863e1e
 github.com/hailocab/go-hostpool e80d13ce29ede4452c43dea11e79b9bc8a15b478
+github.com/hashicorp/consul 5aa90455ce78d4d41578bafc86305e6e6b28d7d2
 github.com/hpcloud/tail b2940955ab8b26e19d43a43c4da0475dd81bdb56
 github.com/influxdata/config b79f6829346b8d6e78ba73544b1e1038f1f1c9da
 github.com/influxdata/influxdb e094138084855d444195b252314dfee9eae34cab

README.md

@@ -145,6 +145,7 @@ Currently implemented sources:
 * [cassandra](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/cassandra)
 * [ceph](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/ceph)
 * [chrony](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/chrony)
+* [consul](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/consul)
 * [conntrack](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/conntrack)
 * [couchbase](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/couchbase)
 * [couchdb](https://github.com/influxdata/telegraf/tree/master/plugins/inputs/couchdb)

plugins/inputs/all/all.go

@@ -9,6 +9,7 @@ import (
 	_ "github.com/influxdata/telegraf/plugins/inputs/chrony"
 	_ "github.com/influxdata/telegraf/plugins/inputs/cloudwatch"
 	_ "github.com/influxdata/telegraf/plugins/inputs/conntrack"
+	_ "github.com/influxdata/telegraf/plugins/inputs/consul"
 	_ "github.com/influxdata/telegraf/plugins/inputs/couchbase"
 	_ "github.com/influxdata/telegraf/plugins/inputs/couchdb"
 	_ "github.com/influxdata/telegraf/plugins/inputs/disque"

plugins/inputs/consul/README.md

@@ -0,0 +1,46 @@
+# Telegraf Input Plugin: Consul
+
+This plugin will collect statistics about all helath checks registered in the Consul. It uses [Consul API](https://www.consul.io/docs/agent/http/health.html#health_state)
+to query the data. It will not report the [telemetry](https://www.consul.io/docs/agent/telemetry.html) but Consul can report those stats already using StatsD protocol if needed.
+
+## Configuration:
+
+```
+# Gather health check statuses from services registered in Consul
+[[inputs.consul]]
+  ## Most of these values defaults to the one configured on a Consul's agent level.
+  ## Optional Consul server address (default: "")
+  # address = ""
+  ## Optional URI scheme for the Consul server (default: "")
+  # scheme = ""
+  ## Optional ACL token used in every request (default: "")
+  # token = ""
+  ## Optional username used for request HTTP Basic Authentication (default: "")
+  # username = ""
+  ## Optional password used for HTTP Basic Authentication (default: "")
+  # password = ""
+  ## Optional data centre to query the health checks from (default: "")
+  # datacentre = ""
+```
+
+## Measurements:
+
+### Consul:
+Tags:
+- node: on which node check/service is registered on
+- service_name: name of the service (this is the service name not the service ID)
+
+Fields:
+- check_id
+- check_name
+- service_id
+- status
+
+## Example output
+
+```
+$ telegraf --config ./telegraf.conf -input-filter consul -test
+* Plugin: consul, Collection 1
+> consul_health_checks,host=wolfpit,node=consul-server-node check_id="serfHealth",check_name="Serf Health Status",service_id="",status="passing" 1464698464486439902
+> consul_health_checks,host=wolfpit,node=consul-server-node,service_name=www.example.com check_id="service:www-example-com.test01",check_name="Service 'www.example.com' check",service_id="www-example-com.test01",status="critical" 1464698464486519036
+```

plugins/inputs/consul/consul.go

@@ -0,0 +1,136 @@
+package consul
+
+import (
+	"net/http"
+
+	"github.com/hashicorp/consul/api"
+	"github.com/influxdata/telegraf"
+	"github.com/influxdata/telegraf/internal"
+	"github.com/influxdata/telegraf/plugins/inputs"
+)
+
+type Consul struct {
+	Address    string
+	Scheme     string
+	Token      string
+	Username   string
+	Password   string
+	Datacentre string
+
+	// Path to CA file
+	SSLCA string `toml:"ssl_ca"`
+	// Path to host cert file
+	SSLCert string `toml:"ssl_cert"`
+	// Path to cert key file
+	SSLKey string `toml:"ssl_key"`
+	// Use SSL but skip chain & host verification
+	InsecureSkipVerify bool
+
+	// client used to connect to Consul agnet
+	client *api.Client
+}
+
+var sampleConfig = `
+  ## Most of these values defaults to the one configured on a Consul's agent level.
+  ## Optional Consul server address (default: "localhost")
+  # address = "localhost"
+  ## Optional URI scheme for the Consul server (default: "http")
+  # scheme = "http"
+  ## Optional ACL token used in every request (default: "")
+  # token = ""
+  ## Optional username used for request HTTP Basic Authentication (default: "")
+  # username = ""
+  ## Optional password used for HTTP Basic Authentication (default: "")
+  # password = ""
+  ## Optional data centre to query the health checks from (default: "")
+  # datacentre = ""
+`
+
+func (c *Consul) Description() string {
+	return "Gather health check statuses from services registered in Consul"
+}
+
+func (c *Consul) SampleConfig() string {
+	return sampleConfig
+}
+
+func (c *Consul) createAPIClient() (*api.Client, error) {
+	config := api.DefaultConfig()
+
+	if c.Address != "" {
+		config.Address = c.Address
+	}
+
+	if c.Scheme != "" {
+		config.Scheme = c.Scheme
+	}
+
+	if c.Datacentre != "" {
+		config.Datacenter = c.Datacentre
+	}
+
+	if c.Username != "" {
+		config.HttpAuth = &api.HttpBasicAuth{
+			Username: c.Username,
+			Password: c.Password,
+		}
+	}
+
+	tlsCfg, err := internal.GetTLSConfig(
+		c.SSLCert, c.SSLKey, c.SSLCA, c.InsecureSkipVerify)
+
+	if err != nil {
+		return nil, err
+	}
+
+	config.HttpClient.Transport = &http.Transport{
+		TLSClientConfig: tlsCfg,
+	}
+
+	return api.NewClient(config)
+}
+
+func (c *Consul) GatherHealthCheck(acc telegraf.Accumulator, checks []*api.HealthCheck) {
+	for _, check := range checks {
+		record := make(map[string]interface{})
+		tags := make(map[string]string)
+
+		record["check_id"] = check.CheckID
+		record["check_name"] = check.Name
+		record["service_id"] = check.ServiceID
+		record["status"] = check.Status
+
+		tags["node"] = check.Node
+		tags["service_name"] = check.ServiceName
+
+		acc.AddFields("consul_health_checks", record, tags)
+	}
+}
+
+func (c *Consul) Gather(acc telegraf.Accumulator) error {
+	if c.client == nil {
+		newClient, err := c.createAPIClient()
+
+		if err != nil {
+			return err
+		}
+
+		c.client = newClient
+	}
+
+	checks, _, err := c.client.Health().State("any", nil)
+
+	if err != nil {
+		return err
+	}
+
+	c.GatherHealthCheck(acc, checks)
+
+	return nil
+}
+
+func init() {
+	inputs.Add("consul", func() telegraf.Input {
+		return &Consul{}
+	})
+}

plugins/inputs/consul/consul_test.go

@@ -0,0 +1,42 @@
+package consul
+
+import (
+	"testing"
+
+	"github.com/hashicorp/consul/api"
+	"github.com/influxdata/telegraf/testutil"
+)
+
+var sampleChecks = []*api.HealthCheck{
+	&api.HealthCheck{
+		Node:        "localhost",
+		CheckID:     "foo.health123",
+		Name:        "foo.health",
+		Status:      "passing",
+		Notes:       "lorem ipsum",
+		Output:      "OK",
+		ServiceID:   "foo.123",
+		ServiceName: "foo",
+	},
+}
+
+func TestGatherHealtCheck(t *testing.T) {
+	expectedFields := map[string]interface{}{
+		"check_id":   "foo.health123",
+		"check_name": "foo.health",
+		"status":     "passing",
+		"service_id": "foo.123",
+	}
+
+	expectedTags := map[string]string{
+		"node":         "localhost",
+		"service_name": "foo",
+	}
+
+	var acc testutil.Accumulator
+
+	consul := &Consul{}
+	consul.GatherHealthCheck(&acc, sampleChecks)
+
+	acc.AssertContainsTaggedFields(t, "consul_health_checks", expectedFields, expectedTags)
+}