Merge pull request #2 from rtkkroland/ee_ccs

[AWN-42236] Add remote info to Elasticsearch exporter

Author: rtkkroland

ccscconsole-i-0408d856fa68f2eee.ccsc.gjacobs.dev2.awn

@@ -0,0 +1,165 @@
+package collector
+
+import (
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"net/url"
+	"path"
+
+	"github.com/go-kit/kit/log"
+	"github.com/go-kit/kit/log/level"
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+// Labels for remote info metrics
+var defaulRemoteInfoLabels = []string{"remote_cluster"}
+var defaultRemoteInfoLabelValues = func(remote_cluster string) []string {
+	return []string{
+		remote_cluster,
+	}
+}
+
+type remoteInfoMetric struct {
+	Type   prometheus.ValueType
+	Desc   *prometheus.Desc
+	Value  func(remoteStats RemoteCluster) float64
+	Labels func(remote_cluster string) []string
+}
+
+// RemoteInfo information struct
+type RemoteInfo struct {
+	logger log.Logger
+	client *http.Client
+	url    *url.URL
+
+	up                              prometheus.Gauge
+	totalScrapes, jsonParseFailures prometheus.Counter
+
+	remoteInfoMetrics []*remoteInfoMetric
+}
+
+// NewClusterSettings defines Cluster Settings Prometheus metrics
+func NewRemoteInfo(logger log.Logger, client *http.Client, url *url.URL) *RemoteInfo {
+
+	return &RemoteInfo{
+		logger: logger,
+		client: client,
+		url:    url,
+
+		up: prometheus.NewGauge(prometheus.GaugeOpts{
+			Name: prometheus.BuildFQName(namespace, "remote_info_stats", "up"),
+			Help: "Was the last scrape of the ElasticSearch remote info endpoint successful.",
+		}),
+		totalScrapes: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: prometheus.BuildFQName(namespace, "remote_info_stats", "total_scrapes"),
+			Help: "Current total ElasticSearch remote info scrapes.",
+		}),
+		jsonParseFailures: prometheus.NewCounter(prometheus.CounterOpts{
+			Name: prometheus.BuildFQName(namespace, "remote_info_stats", "json_parse_failures"),
+			Help: "Number of errors while parsing JSON.",
+		}),
+		// Send all of the remote metrics
+		remoteInfoMetrics: []*remoteInfoMetric{
+			{
+				Type: prometheus.GaugeValue,
+				Desc: prometheus.NewDesc(
+					prometheus.BuildFQName(namespace, "remote_info", "num_nodes_connected"),
+					"Number of nodes connected", defaulRemoteInfoLabels, nil,
+				),
+				Value: func(remoteStats RemoteCluster) float64 {
+					return float64(remoteStats.NumNodesConnected)
+				},
+				Labels: defaultRemoteInfoLabelValues,
+			},
+			{
+				Type: prometheus.GaugeValue,
+				Desc: prometheus.NewDesc(
+					prometheus.BuildFQName(namespace, "remote_info", "max_connections_per_cluster"),
+					"Max connections per cluster", defaulRemoteInfoLabels, nil,
+				),
+				Value: func(remoteStats RemoteCluster) float64 {
+					return float64(remoteStats.MaxConnectionsPerCluster)
+				},
+				Labels: defaultRemoteInfoLabelValues,
+			},
+		},
+	}
+}
+
+func (c *RemoteInfo) fetchAndDecodeRemoteInfoStats() (RemoteInfoResponse, error) {
+	var rir RemoteInfoResponse
+
+	u := *c.url
+	u.Path = path.Join(u.Path, "/_remote/info")
+
+	res, err := c.client.Get(u.String())
+	if err != nil {
+		return rir, fmt.Errorf("failed to get remote info from %s://%s:%s%s: %s",
+			u.Scheme, u.Hostname(), u.Port(), u.Path, err)
+	}
+
+	defer func() {
+		err = res.Body.Close()
+		if err != nil {
+			_ = level.Warn(c.logger).Log(
+				"msg", "failed to close http.Client",
+				"err", err,
+			)
+		}
+	}()
+
+	if res.StatusCode != http.StatusOK {
+		return rir, fmt.Errorf("HTTP Request failed with code %d", res.StatusCode)
+	}
+
+	if err := json.NewDecoder(res.Body).Decode(&rir); err != nil {
+		c.jsonParseFailures.Inc()
+		return rir, err
+	}
+	return rir, nil
+}
+
+// Collect gets remote info values
+func (ri *RemoteInfo) Collect(ch chan<- prometheus.Metric) {
+	ri.totalScrapes.Inc()
+	defer func() {
+		ch <- ri.up
+		ch <- ri.totalScrapes
+		ch <- ri.jsonParseFailures
+	}()
+
+	remoteInfoResp, err := ri.fetchAndDecodeRemoteInfoStats()
+	if err != nil {
+		ri.up.Set(0)
+		_ = level.Warn(ri.logger).Log(
+			"msg", "failed to fetch and decode remote info",
+			"err", err,
+		)
+		return
+	}
+	ri.totalScrapes.Inc()
+	ri.up.Set(1)
+
+	// Remote Info
+	for remote_cluster, remoteInfo := range remoteInfoResp {
+		for _, metric := range ri.remoteInfoMetrics {
+			ch <- prometheus.MustNewConstMetric(
+				metric.Desc,
+				metric.Type,
+				metric.Value(remoteInfo),
+				metric.Labels(remote_cluster)...,
+			)
+		}
+	}
+}
+
+// Describe add Indices metrics descriptions
+func (ri *RemoteInfo) Describe(ch chan<- *prometheus.Desc) {
+	for _, metric := range ri.remoteInfoMetrics {
+		ch <- metric.Desc
+	}
+	ch <- ri.up.Desc()
+	ch <- ri.totalScrapes.Desc()
+	ch <- ri.jsonParseFailures.Desc()
+}

collector/remote_info.go

@@ -0,0 +1,14 @@
+package collector
+
+// RemoteInfoResponse is a representation of a Elasticsearch _remote/info
+type RemoteInfoResponse map[string]RemoteCluster
+
+// RemoteClsuter defines the struct of the tree for the Remote Cluster
+type RemoteCluster struct {
+	Seeds                    []string `json:"seeds"`
+	Connected                bool     `json:"connected"`
+	NumNodesConnected        int64    `json:"num_nodes_connected"`
+	MaxConnectionsPerCluster int64    `json:"max_connections_per_cluster"`
+	InitialConnectTimeout    string   `json:"initial_connect_timeout"`
+	SkipUnavailable          bool     `json:"skip_unavailable"`
+}

collector/remote_info_response.go

@@ -0,0 +1,43 @@
+package collector
+
+import (
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"net/url"
+	"os"
+	"testing"
+
+	"github.com/go-kit/kit/log"
+)
+
+func TestRemoteInfoStats(t *testing.T) {
+	// Testcases created using:
+	//  docker run -d -p 9200:9200 elasticsearch:VERSION-alpine
+	//  curl http://localhost:9200/_cluster/settings/?include_defaults=true
+	files := []string{"../fixtures/settings-5.4.2.json", "../fixtures/settings-merge-5.4.2.json"}
+	for _, filename := range files {
+		f, _ := os.Open(filename)
+		defer f.Close()
+		for hn, handler := range map[string]http.Handler{
+			"plain": http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+				io.Copy(w, f)
+			}),
+		} {
+			ts := httptest.NewServer(handler)
+			defer ts.Close()
+
+			u, err := url.Parse(ts.URL)
+			if err != nil {
+				t.Fatalf("Failed to parse URL: %s", err)
+			}
+			c := NewRemoteInfo(log.NewNopLogger(), http.DefaultClient, u)
+			nsr, err := c.fetchAndDecodeRemoteInfoStats()
+			if err != nil {
+				t.Fatalf("Failed to fetch or decode remote info stats: %s", err)
+			}
+			t.Logf("[%s/%s] Remote Info Stats Response: %+v", hn, filename, nsr)
+
+		}
+	}
+}

collector/remote_info_test.go

@@ -7,11 +7,18 @@ require (
 	github.com/blang/semver/v4 v4.0.0
 	github.com/go-kit/kit v0.10.0
 	github.com/golang/protobuf v1.4.2 // indirect
+	github.com/google/go-github/v25 v25.1.3 // indirect
 	github.com/imdario/mergo v0.3.9
-	github.com/prometheus/client_golang v1.6.0
+	github.com/pkg/errors v0.9.1 // indirect
+	github.com/prometheus/client_golang v1.7.1
 	github.com/prometheus/common v0.10.0
+	github.com/prometheus/promu v0.5.0 // indirect
 	github.com/stretchr/testify v1.5.1 // indirect
-	golang.org/x/sys v0.0.0-20200523222454-059865788121 // indirect
+	golang.org/x/net v0.0.0-20200707034311-ab3426394381 // indirect
+	golang.org/x/oauth2 v0.0.0-20200107190931-bf48bf16ab8d // indirect
+	golang.org/x/sys v0.0.0-20200625212154-ddb9806d33ae // indirect
+	google.golang.org/appengine v1.6.6 // indirect
+	google.golang.org/protobuf v1.25.0 // indirect
 	gopkg.in/alecthomas/kingpin.v2 v2.2.6
 	gopkg.in/yaml.v2 v2.3.0 // indirect
 )