chore: test metricscache on postgres (#16711)

DanielleMaywood · web-flow · commit 6dd51f92fbd6 · 2025-02-27T09:43:51.000Z
metricscache_test has been running tests against dbmem only, instead of
against postgres. Unfortunately the implementations of
GetTemplateAverageBuildTime have diverged between dbmem and postgres.
This change gets the tests working on Postgres and test for the
behaviour postgres provides.
diff --git a/coderd/coderd.go b/coderd/coderd.go
@@ -422,6 +422,7 @@ func New(options *Options) *API {
 	metricsCache := metricscache.New(
 		options.Database,
 		options.Logger.Named("metrics_cache"),
+		options.Clock,
 		metricscache.Intervals{
 			TemplateBuildTimes: options.MetricsCacheRefreshInterval,
 			DeploymentStats:    options.AgentStatsRefreshInterval,
diff --git a/coderd/database/dbmem/dbmem.go b/coderd/database/dbmem/dbmem.go
@@ -269,7 +269,7 @@ type data struct {
 	presetParameters                 []database.TemplateVersionPresetParameter
 }
 
-func tryPercentile(fs []float64, p float64) float64 {
+func tryPercentileCont(fs []float64, p float64) float64 {
 	if len(fs) == 0 {
 		return -1
 	}
@@ -282,6 +282,14 @@ func tryPercentile(fs []float64, p float64) float64 {
 	return fs[lower] + (fs[upper]-fs[lower])*(pos-float64(lower))
 }
 
+func tryPercentileDisc(fs []float64, p float64) float64 {
+	if len(fs) == 0 {
+		return -1
+	}
+	sort.Float64s(fs)
+	return fs[max(int(math.Ceil(float64(len(fs))*p/100-1)), 0)]
+}
+
 func validateDatabaseTypeWithValid(v reflect.Value) (handled bool, err error) {
 	if v.Kind() == reflect.Struct {
 		return false, nil
@@ -2790,8 +2798,8 @@ func (q *FakeQuerier) GetDeploymentWorkspaceAgentStats(_ context.Context, create
 		latencies = append(latencies, agentStat.ConnectionMedianLatencyMS)
 	}
 
-	stat.WorkspaceConnectionLatency50 = tryPercentile(latencies, 50)
-	stat.WorkspaceConnectionLatency95 = tryPercentile(latencies, 95)
+	stat.WorkspaceConnectionLatency50 = tryPercentileCont(latencies, 50)
+	stat.WorkspaceConnectionLatency95 = tryPercentileCont(latencies, 95)
 
 	return stat, nil
 }
@@ -2839,8 +2847,8 @@ func (q *FakeQuerier) GetDeploymentWorkspaceAgentUsageStats(_ context.Context, c
 		stat.WorkspaceTxBytes += agentStat.TxBytes
 		latencies = append(latencies, agentStat.ConnectionMedianLatencyMS)
 	}
-	stat.WorkspaceConnectionLatency50 = tryPercentile(latencies, 50)
-	stat.WorkspaceConnectionLatency95 = tryPercentile(latencies, 95)
+	stat.WorkspaceConnectionLatency50 = tryPercentileCont(latencies, 50)
+	stat.WorkspaceConnectionLatency95 = tryPercentileCont(latencies, 95)
 
 	for _, agentStat := range sessions {
 		stat.SessionCountVSCode += agentStat.SessionCountVSCode
@@ -4987,9 +4995,9 @@ func (q *FakeQuerier) GetTemplateAverageBuildTime(ctx context.Context, arg datab
 	}
 
 	var row database.GetTemplateAverageBuildTimeRow
-	row.Delete50, row.Delete95 = tryPercentile(deleteTimes, 50), tryPercentile(deleteTimes, 95)
-	row.Stop50, row.Stop95 = tryPercentile(stopTimes, 50), tryPercentile(stopTimes, 95)
-	row.Start50, row.Start95 = tryPercentile(startTimes, 50), tryPercentile(startTimes, 95)
+	row.Delete50, row.Delete95 = tryPercentileDisc(deleteTimes, 50), tryPercentileDisc(deleteTimes, 95)
+	row.Stop50, row.Stop95 = tryPercentileDisc(stopTimes, 50), tryPercentileDisc(stopTimes, 95)
+	row.Start50, row.Start95 = tryPercentileDisc(startTimes, 50), tryPercentileDisc(startTimes, 95)
 	return row, nil
 }
 
@@ -6024,8 +6032,8 @@ func (q *FakeQuerier) GetUserLatencyInsights(_ context.Context, arg database.Get
 			Username:                     user.Username,
 			AvatarURL:                    user.AvatarURL,
 			TemplateIDs:                  seenTemplatesByUserID[userID],
-			WorkspaceConnectionLatency50: tryPercentile(latencies, 50),
-			WorkspaceConnectionLatency95: tryPercentile(latencies, 95),
+			WorkspaceConnectionLatency50: tryPercentileCont(latencies, 50),
+			WorkspaceConnectionLatency95: tryPercentileCont(latencies, 95),
 		}
 		rows = append(rows, row)
 	}
@@ -6669,8 +6677,8 @@ func (q *FakeQuerier) GetWorkspaceAgentStats(_ context.Context, createdAfter tim
 		if !ok {
 			continue
 		}
-		stat.WorkspaceConnectionLatency50 = tryPercentile(latencies, 50)
-		stat.WorkspaceConnectionLatency95 = tryPercentile(latencies, 95)
+		stat.WorkspaceConnectionLatency50 = tryPercentileCont(latencies, 50)
+		stat.WorkspaceConnectionLatency95 = tryPercentileCont(latencies, 95)
 		statByAgent[stat.AgentID] = stat
 	}
 
@@ -6807,8 +6815,8 @@ func (q *FakeQuerier) GetWorkspaceAgentUsageStats(_ context.Context, createdAt t
 	for key, latencies := range latestAgentLatencies {
 		val, ok := latestAgentStats[key]
 		if ok {
-			val.WorkspaceConnectionLatency50 = tryPercentile(latencies, 50)
-			val.WorkspaceConnectionLatency95 = tryPercentile(latencies, 95)
+			val.WorkspaceConnectionLatency50 = tryPercentileCont(latencies, 50)
+			val.WorkspaceConnectionLatency95 = tryPercentileCont(latencies, 95)
 		}
 		latestAgentStats[key] = val
 	}
diff --git a/coderd/database/queries.sql.go b/coderd/database/queries.sql.go
diff --git a/coderd/database/queries/workspaces.sql b/coderd/database/queries/workspaces.sql
@@ -415,13 +415,11 @@ WHERE
 ORDER BY created_at DESC;
 
 -- name: GetWorkspaceUniqueOwnerCountByTemplateIDs :many
-SELECT
-	template_id, COUNT(DISTINCT owner_id) AS unique_owners_sum
-FROM
-	workspaces
-WHERE
-	template_id = ANY(@template_ids :: uuid[]) AND deleted = false
-GROUP BY template_id;
+SELECT templates.id AS template_id, COUNT(DISTINCT workspaces.owner_id) AS unique_owners_sum
+FROM templates
+LEFT JOIN workspaces ON workspaces.template_id = templates.id AND workspaces.deleted = false
+WHERE templates.id = ANY(@template_ids :: uuid[])
+GROUP BY templates.id;
 
 -- name: InsertWorkspace :one
 INSERT INTO
diff --git a/coderd/metricscache/metricscache.go b/coderd/metricscache/metricscache.go
@@ -15,6 +15,7 @@ import (
 	"github.com/coder/coder/v2/coderd/database/dbauthz"
 	"github.com/coder/coder/v2/coderd/database/dbtime"
 	"github.com/coder/coder/v2/codersdk"
+	"github.com/coder/quartz"
 	"github.com/coder/retry"
 )
 
@@ -26,6 +27,7 @@ import (
 type Cache struct {
 	database  database.Store
 	log       slog.Logger
+	clock     quartz.Clock
 	intervals Intervals
 
 	templateWorkspaceOwners  atomic.Pointer[map[uuid.UUID]int]
@@ -45,7 +47,7 @@ type Intervals struct {
 	DeploymentStats    time.Duration
 }
 
-func New(db database.Store, log slog.Logger, intervals Intervals, usage bool) *Cache {
+func New(db database.Store, log slog.Logger, clock quartz.Clock, intervals Intervals, usage bool) *Cache {
 	if intervals.TemplateBuildTimes <= 0 {
 		intervals.TemplateBuildTimes = time.Hour
 	}
@@ -55,6 +57,7 @@ func New(db database.Store, log slog.Logger, intervals Intervals, usage bool) *C
 	ctx, cancel := context.WithCancel(context.Background())
 
 	c := &Cache{
+		clock:     clock,
 		database:  db,
 		intervals: intervals,
 		log:       log,
@@ -104,7 +107,7 @@ func (c *Cache) refreshTemplateBuildTimes(ctx context.Context) error {
 				Valid: true,
 			},
 			StartTime: sql.NullTime{
-				Time:  dbtime.Time(time.Now().AddDate(0, 0, -30)),
+				Time:  dbtime.Time(c.clock.Now().AddDate(0, 0, -30)),
 				Valid: true,
 			},
 		})
@@ -131,7 +134,7 @@ func (c *Cache) refreshTemplateBuildTimes(ctx context.Context) error {
 
 func (c *Cache) refreshDeploymentStats(ctx context.Context) error {
 	var (
-		from       = dbtime.Now().Add(-15 * time.Minute)
+		from       = c.clock.Now().Add(-15 * time.Minute)
 		agentStats database.GetDeploymentWorkspaceAgentStatsRow
 		err        error
 	)
@@ -155,8 +158,8 @@ func (c *Cache) refreshDeploymentStats(ctx context.Context) error {
 	}
 	c.deploymentStatsResponse.Store(&codersdk.DeploymentStats{
 		AggregatedFrom: from,
-		CollectedAt:    dbtime.Now(),
-		NextUpdateAt:   dbtime.Now().Add(c.intervals.DeploymentStats),
+		CollectedAt:    dbtime.Time(c.clock.Now()),
+		NextUpdateAt:   dbtime.Time(c.clock.Now().Add(c.intervals.DeploymentStats)),
 		Workspaces: codersdk.WorkspaceDeploymentStats{
 			Pending:  workspaceStats.PendingWorkspaces,
 			Building: workspaceStats.BuildingWorkspaces,
diff --git a/coderd/metricscache/metricscache_test.go b/coderd/metricscache/metricscache_test.go
