feat(metrics): expose velero_build_info gauge

changelogs/unreleased/9934-amarkdotdev

@@ -0,0 +1 @@
+Add velero_build_info metric exposing the Velero version, git SHA, git tree state, and Go runtime details

pkg/cmd/server/server.go

@@ -23,6 +23,7 @@ import (
 	"net/http"
 	"net/http/pprof"
 	"os"
+	goruntime "runtime"
 	"strings"
 	"time"
 
@@ -560,6 +561,7 @@ func (s *server) runControllers(defaultVolumeSnapshotLocations map[string]string
 	}()
 	s.metrics = metrics.NewServerMetrics()
 	s.metrics.RegisterAllMetrics()
+	registerBuildInfo(s.metrics)
 	// Initialize manual backup metrics
 	s.metrics.InitSchedule("")
 
@@ -1204,3 +1206,14 @@ func markPodVolumeRestoresCancel(ctx context.Context, client ctrlclient.Client,
 		}
 	}
 }
+
+func registerBuildInfo(m *metrics.ServerMetrics) {
+	m.RegisterBuildInfo(
+		buildinfo.Version,
+		buildinfo.GitSHA,
+		buildinfo.GitTreeState,
+		goruntime.Version(),
+		goruntime.GOOS,
+		goruntime.GOARCH,
+	)
+}

pkg/cmd/server/server_test.go

@@ -24,6 +24,8 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/mock"
 	"github.com/stretchr/testify/require"
+	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/client_golang/prometheus"
 	corev1api "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
@@ -34,10 +36,12 @@ import (
 	velerov1api "github.com/vmware-tanzu/velero/pkg/apis/velero/v1"
 	velerov2alpha1api "github.com/vmware-tanzu/velero/pkg/apis/velero/v2alpha1"
 	"github.com/vmware-tanzu/velero/pkg/builder"
+	"github.com/vmware-tanzu/velero/pkg/buildinfo"
 	"github.com/vmware-tanzu/velero/pkg/client/mocks"
 	"github.com/vmware-tanzu/velero/pkg/cmd/server/config"
 	"github.com/vmware-tanzu/velero/pkg/constant"
 	discovery_mocks "github.com/vmware-tanzu/velero/pkg/discovery/mocks"
+	"github.com/vmware-tanzu/velero/pkg/metrics"
 	velerotest "github.com/vmware-tanzu/velero/pkg/test"
 	"github.com/vmware-tanzu/velero/pkg/uploader"
 )
@@ -448,3 +452,31 @@ func Test_setDefaultBackupLocation(t *testing.T) {
 	err = setDefaultBackupLocation(t.Context(), c, "velero", "default", logrus.New())
 	assert.NoError(t, err)
 }
+
+func TestRegisterBuildInfo(t *testing.T) {
+	m := metrics.NewServerMetrics()
+	m.RegisterAllMetrics()
+	registerBuildInfo(m)
+
+	ch := make(chan prometheus.Metric, 1)
+	m.Metrics()["build_info"].(*prometheus.GaugeVec).Collect(ch)
+	close(ch)
+
+	var found bool
+	for metric := range ch {
+		dtoMetric := &dto.Metric{}
+		require.NoError(t, metric.Write(dtoMetric))
+		if dtoMetric.GetGauge().GetValue() != 1 {
+			continue
+		}
+		labels := map[string]string{}
+		for _, label := range dtoMetric.Label {
+			labels[label.GetName()] = label.GetValue()
+		}
+		assert.Equal(t, buildinfo.Version, labels["version"])
+		assert.Equal(t, buildinfo.GitSHA, labels["git_commit"])
+		assert.Equal(t, buildinfo.GitTreeState, labels["git_tree_state"])
+		found = true
+	}
+	require.True(t, found, "build_info metric not found")
+}

pkg/metrics/metrics.go

@@ -90,6 +90,7 @@ const (
 	// Each completed job's duration is recorded in the appropriate bucket, allowing
 	// analysis of individual job performance and trending over time.
 	repoMaintenanceDurationSeconds = "repo_maintenance_duration_seconds"
+	buildInfo                      = "build_info"
 
 	// Labels
 	nodeMetricLabel         = "node"
@@ -393,6 +394,14 @@ func NewServerMetrics() *ServerMetrics {
 				},
 				[]string{repositoryNameLabel},
 			),
+			buildInfo: prometheus.NewGaugeVec(
+				prometheus.GaugeOpts{
+					Namespace: metricNamespace,
+					Name:      buildInfo,
+					Help:      "A metric with a constant '1' value labeled with Velero build information.",
+				},
+				[]string{"version", "git_commit", "git_tree_state", "go_version", "goos", "goarch"},
+			),
 		},
 	}
 }
@@ -502,6 +511,13 @@ func (m *ServerMetrics) RegisterAllMetrics() {
 	}
 }
 
+// RegisterBuildInfo records static build information for the Velero server.
+func (m *ServerMetrics) RegisterBuildInfo(version, gitCommit, gitTreeState, goVersion, goos, goarch string) {
+	if g, ok := m.metrics[buildInfo].(*prometheus.GaugeVec); ok {
+		g.WithLabelValues(version, gitCommit, gitTreeState, goVersion, goos, goarch).Set(1)
+	}
+}
+
 // InitSchedule initializes counter metrics of a schedule.
 func (m *ServerMetrics) InitSchedule(scheduleName string) {
 	if c, ok := m.metrics[backupAttemptTotal].(*prometheus.CounterVec); ok {

pkg/metrics/metrics_test.go

@@ -601,3 +601,35 @@ func getMaintenanceHistogramCount(t *testing.T, vec *prometheus.HistogramVec, re
 	t.Fatalf("Histogram with repository_name label '%s' not found", repositoryName)
 	return 0
 }
+
+func TestRegisterBuildInfo(t *testing.T) {
+	m := NewServerMetrics()
+	m.RegisterBuildInfo("v1.13.0", "abc123", "clean", "go1.22.0", "linux", "amd64")
+
+	ch := make(chan prometheus.Metric, 1)
+	m.metrics[buildInfo].(*prometheus.GaugeVec).Collect(ch)
+	close(ch)
+
+	var found bool
+	for metric := range ch {
+		dtoMetric := &dto.Metric{}
+		require.NoError(t, metric.Write(dtoMetric))
+		if dtoMetric.GetGauge().GetValue() != 1 {
+			continue
+		}
+		labels := map[string]string{}
+		for _, label := range dtoMetric.Label {
+			labels[label.GetName()] = label.GetValue()
+		}
+		assert.Equal(t, map[string]string{
+			"version":        "v1.13.0",
+			"git_commit":     "abc123",
+			"git_tree_state": "clean",
+			"go_version":     "go1.22.0",
+			"goos":           "linux",
+			"goarch":         "amd64",
+		}, labels)
+		found = true
+	}
+	require.True(t, found, "build_info metric not found")
+}