add build_info metrics

Signed-off-by: yeya24 <[email protected]>

dfdaemon/server.go

@@ -9,9 +9,10 @@ import (
 	"github.com/dragonflyoss/Dragonfly/dfdaemon/config"
 	"github.com/dragonflyoss/Dragonfly/dfdaemon/handler"
 	"github.com/dragonflyoss/Dragonfly/dfdaemon/proxy"
-	"github.com/sirupsen/logrus"
+	"github.com/dragonflyoss/Dragonfly/version"
 
 	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
 )
 
 // Server represents the dfdaemon server
@@ -104,6 +105,8 @@ func (s *Server) Start() error {
 	} else {
 		logrus.Infof("start dfdaemon http server on %s", s.server.Addr)
 	}
+	// register dfdaemon build information
+	version.NewBuildInfo("dfdaemon")
 	return s.server.ListenAndServe()
 }
 

docs/user_guide/metrics.md

@@ -0,0 +1,19 @@
+# Prometheus Metrics
+
+This doc contains all the metrics that Dragonfly components currently support. Now we only support metrics for Dfdaemon and SuperNode. And we will support dfget metrics in the future. For Dfdaemon and SuperNode, the metrics path is fixed to /metrics. The following metrics are exported.
+
+## Supernode
+
+- dragonfly_supernode_build_info{version, revision, goversion, arch, os} - build and version information of supernode
+- dragonfly_supernode_http_requests_total{code, handler, method} - total number of http requests
+- dragonfly_supernode_http_request_duration_seconds{code, handler, method} - http request latency in seconds
+- dragonfly_supernode_http_request_size_bytes{code, handler, method} - http request size in bytes
+- dragonfly_supernode_http_response_size_bytes{code, handler, method} - http response size in bytes
+
+## Dfdaemon
+
+- dragonfly_dfdaemon_build_info{version, revision, goversion, arch, os} - build and version information of dfdaemon
+
+## Dfget
+
+TODO

docs/user_guide/monitoring.md

@@ -2,7 +2,7 @@
 
 Currently metrics become an important part of observability. As for monitoring Dragonfly, we recommend you to use Prometheus.
 
-In Dragonfly project, there are two long-running process: supernode and dfdaemon. Both of these two components expose its metrics via `/metrics` endpoint, so Prometheus can get metrics from these two components. We will also support dfget metrics in the future. As for current metrics, you can check out this docs.
+In Dragonfly project, there are two long-running processes: supernode and dfdaemon. Each of the components expose its metrics via `/metrics` endpoint, so Prometheus can get metrics from each component. We will also support dfget metrics in the future. As for current metrics, you can check out [metrics](metrics.md).
 
 ## How to set up Prometheus
 

supernode/server/metrics.go

@@ -43,8 +43,8 @@ func newMetrics() *metrics {
 			prometheus.HistogramOpts{
 				Namespace: config.Namespace,
 				Subsystem: config.Subsystem,
-				Name: "http_request_size_bytes",
-				Help: "Histogram of request size for HTTP requests.",
+				Name:      "http_request_size_bytes",
+				Help:      "Histogram of request size for HTTP requests.",
 				Buckets:   prometheus.ExponentialBuckets(100, 10, 8),
 			},
 			[]string{"code", "handler", "method"},

supernode/server/router_test.go

@@ -2,7 +2,6 @@ package server
 
 import (
 	"encoding/json"
-	"github.com/gorilla/mux"
 	"io/ioutil"
 	"math/rand"
 	"net"
@@ -18,6 +17,7 @@ import (
 	"github.com/dragonflyoss/Dragonfly/version"
 
 	"github.com/go-check/check"
+	"github.com/gorilla/mux"
 	prom_testutil "github.com/prometheus/client_golang/prometheus/testutil"
 )
 
@@ -122,13 +122,13 @@ func (rs *RouterTestSuite) TestHTTPMetrics(c *check.C) {
 
 	counter := m.requestCounter
 	c.Assert(1, check.Equals,
-		int(prom_testutil.ToFloat64(counter.WithLabelValues("/metrics", strconv.Itoa(http.StatusOK)))))
+		int(prom_testutil.ToFloat64(counter.WithLabelValues(strconv.Itoa(http.StatusOK), "/metrics", "get"))))
 
 	for i := 0; i < 5; i++ {
 		code, _, err := cutil.Get("http://"+rs.addr+"/_ping", 0)
 		c.Check(err, check.IsNil)
 		c.Assert(code, check.Equals, 200)
 		c.Assert(i+1, check.Equals,
-			int(prom_testutil.ToFloat64(counter.WithLabelValues("/_ping", strconv.Itoa(http.StatusOK)))))
+			int(prom_testutil.ToFloat64(counter.WithLabelValues(strconv.Itoa(http.StatusOK), "/_ping", "get"))))
 	}
 }

supernode/server/server.go

@@ -15,11 +15,12 @@ import (
 	"github.com/dragonflyoss/Dragonfly/supernode/daemon/mgr/scheduler"
 	"github.com/dragonflyoss/Dragonfly/supernode/daemon/mgr/task"
 	"github.com/dragonflyoss/Dragonfly/supernode/store"
+	"github.com/dragonflyoss/Dragonfly/version"
 
 	"github.com/sirupsen/logrus"
 )
 
-// Server is server instance.
+// Server is supernode server struct.
 type Server struct {
 	Config       *config.Config
 	PeerMgr      mgr.PeerMgr
@@ -78,7 +79,7 @@ func New(cfg *config.Config) (*Server, error) {
 	}, nil
 }
 
-// Start runs
+// Start runs supernode server.
 func (s *Server) Start() error {
 	router := initRoute(s)
 
@@ -90,6 +91,8 @@ func (s *Server) Start() error {
 		return err
 	}
 
+	// register supernode build information
+	version.NewBuildInfo("supernode")
 	server := &http.Server{
 		Handler:           router,
 		ReadTimeout:       time.Minute * 10,

test/api_metrics_test.go

@@ -51,22 +51,21 @@ func (s *APIMetricsSuite) TestHttpMetrics(c *check.C) {
 	CheckRespStatus(c, resp, 200)
 
 	// Get httpRequest counter value equals 1.
-	pingTimes, found := GetMetricValue(c, fmt.Sprintf(requestCounter, 200, "/_ping", "get"))
-	c.Assert(found, check.Equals, true)
-	c.Assert(pingTimes, check.Equals, float64(1))
+	CheckMetric(c, fmt.Sprintf(requestCounter, 200, "/_ping", "get"), 1)
 
 	// Get httpResponse size sum value equals 2.
-	responseBytes, found := GetMetricValue(c, fmt.Sprintf(responseSizeSum, 200, "/_ping", "get"))
-	c.Assert(found, check.Equals, true)
-	c.Assert(responseBytes, check.Equals, float64(2))
+	CheckMetric(c, fmt.Sprintf(responseSizeSum, 200, "/_ping", "get"), 2)
 
 	// Get httpResponse size count value equals 1.
-	responseCount, found := GetMetricValue(c, fmt.Sprintf(responseSizeCount, 200, "/_ping", "get"))
-	c.Assert(found, check.Equals, true)
-	c.Assert(responseCount, check.Equals, float64(1))
+	CheckMetric(c, fmt.Sprintf(responseSizeCount, 200, "/_ping", "get"), 1)
 
 	// Get httpRequest size count value equals 1.
-	requestCount, found := GetMetricValue(c, fmt.Sprintf(requestSizeCount, 200, "/_ping", "get"))
-	c.Assert(found, check.Equals, true)
-	c.Assert(requestCount, check.Equals, float64(1))
+	CheckMetric(c, fmt.Sprintf(requestSizeCount, 200, "/_ping", "get"), 1)
+}
+
+// TestBuildInfoMetrics tests build info metric.
+func (s *APIMetricsSuite) TestBuildInfoMetrics(c *check.C) {
+	supernodeBuildInfo := `dragonfly_supernode_build_info{`
+	// Ensure build_info metric exists.
+	CheckMetric(c, supernodeBuildInfo, 1)
 }

test/util_api.go

@@ -20,9 +20,9 @@ func CheckRespStatus(c *check.C, resp *http.Response, status int) {
 	}
 }
 
-// GetMetricValue get the specific metrics from /metrics endpoint,
-// if cannot find this metrics, return not found.
-func GetMetricValue(c *check.C, key string) (float64, bool) {
+// CheckMetric find the specific metric from /metrics endpoint and it will compare the metric
+// value with expected value.
+func CheckMetric(c *check.C, metric string, value float64) {
 	var val float64
 	resp, err := request.Get("/metrics")
 	c.Assert(err, check.IsNil)
@@ -31,12 +31,14 @@ func GetMetricValue(c *check.C, key string) (float64, bool) {
 	c.Assert(err, check.IsNil)
 	lines := strings.Split(string(data), "\n")
 	for _, line := range lines {
-		if strings.Contains(line, key) {
+		if strings.Contains(line, metric) {
 			val, err = strconv.ParseFloat(strings.Split(line, " ")[1], 64)
 			c.Assert(err, check.IsNil)
-			return val, true
+			c.Assert(val, check.Equals, value)
+			return
 		}
 	}
 
-	return val, false
+	// Cannot find expected metric and fail the test.
+	c.FailNow()
 }

version/version.go

@@ -27,6 +27,9 @@ import (
 	"strings"
 	"text/template"
 
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+
 	"github.com/dragonflyoss/Dragonfly/apis/types"
 )
 
@@ -104,6 +107,23 @@ func Print(program string) string {
 	return strings.TrimSpace(buf.String())
 }
 
+// NewBuildInfo register a collector which exports metrics about version and build information.
+func NewBuildInfo(program string) {
+	buildInfo := promauto.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Namespace: "dragonfly",
+			Subsystem: program,
+			Name:      "build_info",
+			Help: fmt.Sprintf(
+				"A metric with a constant '1' value labeled by version, revision, os, arch and goversion from which %s was built.",
+				program,
+			),
+		},
+		[]string{"version", "revision", "os", "arch", "goversion"},
+	)
+	buildInfo.WithLabelValues(version, revision, os, arch, goVersion).Set(1)
+}
+
 // Handler returns build information
 func Handler(w http.ResponseWriter, r *http.Request) {
 	data, err := json.Marshal(DFVersion)