coder · kylecarbs · Aug 9, 2022 · Aug 8, 2022
diff --git a/cli/server.go b/cli/server.go
@@ -402,11 +402,17 @@ func server() *cobra.Command {
 			}
 			if promEnabled {
 				options.PrometheusRegistry = prometheus.NewRegistry()
-				closeFunc, err := prometheusmetrics.ActiveUsers(ctx, options.PrometheusRegistry, options.Database, 0)
+				closeUsersFunc, err := prometheusmetrics.ActiveUsers(ctx, options.PrometheusRegistry, options.Database, 0)
 				if err != nil {
 					return xerrors.Errorf("register active users prometheus metric: %w", err)
 				}
-				defer closeFunc()
+				defer closeUsersFunc()
+
+				closeWorkspacesFunc, err := prometheusmetrics.Workspaces(ctx, options.PrometheusRegistry, options.Database, 0)
+				if err != nil {
+					return xerrors.Errorf("register workspaces prometheus metric: %w", err)
+				}
+				defer closeWorkspacesFunc()
 
 				//nolint:revive
 				defer serveHandler(ctx, logger, promhttp.InstrumentMetricHandler(

diff --git a/cli/server_test.go b/cli/server_test.go
@@ -416,16 +416,23 @@ func TestServer(t *testing.T) {
 
 		scanner := bufio.NewScanner(res.Body)
 		hasActiveUsers := false
+		hasWorkspaces := false
 		for scanner.Scan() {
 			// This metric is manually registered to be tracked in the server. That's
 			// why we test it's tracked here.
 			if strings.HasPrefix(scanner.Text(), "coderd_api_active_users_duration_hour") {
 				hasActiveUsers = true
 				continue
 			}
+			if strings.HasPrefix(scanner.Text(), "coderd_api_workspace_latest_build_total") {
+				hasWorkspaces = true
+				continue
+			}
+			t.Logf("scanned %s", scanner.Text())
 		}
 		require.NoError(t, scanner.Err())
 		require.True(t, hasActiveUsers)
+		require.True(t, hasWorkspaces)
 		cancelFunc()
 		<-serverErr
 	})

diff --git a/coderd/database/databasefake/databasefake.go b/coderd/database/databasefake/databasefake.go
@@ -600,6 +600,32 @@ func (q *fakeQuerier) GetLatestWorkspaceBuildByWorkspaceID(_ context.Context, wo
 	return row, nil
 }
 
+func (q *fakeQuerier) GetLatestWorkspaceBuilds(_ context.Context) ([]database.WorkspaceBuild, error) {
+	q.mutex.RLock()
+	defer q.mutex.RUnlock()
+
+	builds := make(map[uuid.UUID]database.WorkspaceBuild)
+	buildNumbers := make(map[uuid.UUID]int32)
+	for _, workspaceBuild := range q.workspaceBuilds {
+		id := workspaceBuild.WorkspaceID
+		if workspaceBuild.BuildNumber > buildNumbers[id] {
+			builds[id] = workspaceBuild
+			buildNumbers[id] = workspaceBuild.BuildNumber
+		}
+	}
+	var returnBuilds []database.WorkspaceBuild
+	for i, n := range buildNumbers {
+		if n > 0 {
+			b := builds[i]
+			returnBuilds = append(returnBuilds, b)
+		}
+	}
+	if len(returnBuilds) == 0 {
+		return nil, sql.ErrNoRows
+	}
+	return returnBuilds, nil
+}
+
 func (q *fakeQuerier) GetLatestWorkspaceBuildsByWorkspaceIDs(_ context.Context, ids []uuid.UUID) ([]database.WorkspaceBuild, error) {
 	q.mutex.RLock()
 	defer q.mutex.RUnlock()

diff --git a/coderd/database/modelmethods.go b/coderd/database/modelmethods.go
@@ -1,6 +1,8 @@
 package database
 
-import "github.com/coder/coder/coderd/rbac"
+import (
+	"github.com/coder/coder/coderd/rbac"
+)
 
 func (t Template) RBACObject() rbac.Object {
 	return rbac.ResourceTemplate.InOrg(t.OrganizationID).WithID(t.ID.String())

diff --git a/coderd/database/querier.go b/coderd/database/querier.go
diff --git a/coderd/database/queries.sql.go b/coderd/database/queries.sql.go
diff --git a/coderd/database/queries/workspacebuilds.sql b/coderd/database/queries/workspacebuilds.sql
@@ -99,6 +99,19 @@ JOIN
     workspace_builds wb
 ON m.workspace_id = wb.workspace_id AND m.max_build_number = wb.build_number;
 
+-- name: GetLatestWorkspaceBuilds :many
+SELECT wb.*
+FROM (
+    SELECT
+        workspace_id, MAX(build_number) as max_build_number
+    FROM
+        workspace_builds
+    GROUP BY
+        workspace_id
+) m
+JOIN
+    workspace_builds wb
+ON m.workspace_id = wb.workspace_id AND m.max_build_number = wb.build_number;
 
 -- name: InsertWorkspaceBuild :one
 INSERT INTO

diff --git a/coderd/prometheusmetrics/prometheusmetrics.go b/coderd/prometheusmetrics/prometheusmetrics.go
@@ -7,6 +7,7 @@ import (
 	"github.com/google/uuid"
 	"github.com/prometheus/client_golang/prometheus"
 
+	"github.com/coder/coder/coderd"
 	"github.com/coder/coder/coderd/database"
 )
 
@@ -50,3 +51,56 @@ func ActiveUsers(ctx context.Context, registerer prometheus.Registerer, db datab
 	}()
 	return cancelFunc, nil
 }
+
+// Workspaces tracks the total number of workspaces with labels on status.
+func Workspaces(ctx context.Context, registerer prometheus.Registerer, db database.Store, duration time.Duration) (context.CancelFunc, error) {
+	if duration == 0 {
+		duration = 5 * time.Minute
+	}
+
+	gauge := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: "coderd",
+		Subsystem: "api",
+		Name:      "workspace_latest_build_total",
+		Help:      "The latest workspace builds with a status.",
+	}, []string{"status"})
+	err := registerer.Register(gauge)
+	if err != nil {
+		return nil, err
+	}
+	// This exists so the prometheus metric exports immediately when set.
+	// It helps with tests so they don't have to wait for a tick.
+	gauge.WithLabelValues("pending").Set(0)
+
+	ctx, cancelFunc := context.WithCancel(ctx)
+	ticker := time.NewTicker(duration)
+	go func() {
+		defer ticker.Stop()
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case <-ticker.C:
+			}
+			builds, err := db.GetLatestWorkspaceBuilds(ctx)
+			if err != nil {
+				continue
+			}
+			jobIDs := make([]uuid.UUID, 0, len(builds))
+			for _, build := range builds {
+				jobIDs = append(jobIDs, build.JobID)
+			}
+			jobs, err := db.GetProvisionerJobsByIDs(ctx, jobIDs)
+			if err != nil {
+				continue
+			}
+
+			gauge.Reset()
+			for _, job := range jobs {
+				status := coderd.ConvertProvisionerJobStatus(job)
+				gauge.WithLabelValues(string(status)).Add(1)
+			}
+		}
+	}()
+	return cancelFunc, nil
+}