chore: add more pprof labels for various go routines (#19243)

Emyrk · web-flow · commit 0a3afeddc8a0 · 2025-08-07T20:05:32.000Z
- ReplicaSync
- Notifications
- MetricsAggregator
- DBPurge
diff --git a/coderd/database/dbpurge/dbpurge.go b/coderd/database/dbpurge/dbpurge.go
@@ -12,6 +12,7 @@ import (
 	"github.com/coder/coder/v2/coderd/database"
 	"github.com/coder/coder/v2/coderd/database/dbauthz"
 	"github.com/coder/coder/v2/coderd/database/dbtime"
+	"github.com/coder/coder/v2/coderd/pproflabel"
 	"github.com/coder/quartz"
 )
 
@@ -38,7 +39,7 @@ func New(ctx context.Context, logger slog.Logger, db database.Store, clk quartz.
 
 	// Start the ticker with the initial delay.
 	ticker := clk.NewTicker(delay)
-	doTick := func(start time.Time) {
+	doTick := func(ctx context.Context, start time.Time) {
 		defer ticker.Reset(delay)
 		// Start a transaction to grab advisory lock, we don't want to run
 		// multiple purges at the same time (multiple replicas).
@@ -85,21 +86,21 @@ func New(ctx context.Context, logger slog.Logger, db database.Store, clk quartz.
 		}
 	}
 
-	go func() {
+	pproflabel.Go(ctx, pproflabel.Service(pproflabel.ServiceDBPurge), func(ctx context.Context) {
 		defer close(closed)
 		defer ticker.Stop()
 		// Force an initial tick.
-		doTick(dbtime.Time(clk.Now()).UTC())
+		doTick(ctx, dbtime.Time(clk.Now()).UTC())
 		for {
 			select {
 			case <-ctx.Done():
 				return
 			case tick := <-ticker.C:
 				ticker.Stop()
-				doTick(dbtime.Time(tick).UTC())
+				doTick(ctx, dbtime.Time(tick).UTC())
 			}
 		}
-	}()
+	})
 	return &instance{
 		cancel: cancelFunc,
 		closed: closed,
diff --git a/coderd/notifications/manager.go b/coderd/notifications/manager.go
@@ -11,12 +11,13 @@ import (
 	"golang.org/x/xerrors"
 
 	"cdr.dev/slog"
-	"github.com/coder/quartz"
 
 	"github.com/coder/coder/v2/coderd/database"
 	"github.com/coder/coder/v2/coderd/database/pubsub"
 	"github.com/coder/coder/v2/coderd/notifications/dispatch"
+	"github.com/coder/coder/v2/coderd/pproflabel"
 	"github.com/coder/coder/v2/codersdk"
+	"github.com/coder/quartz"
 )
 
 var ErrInvalidDispatchTimeout = xerrors.New("dispatch timeout must be less than lease period")
@@ -145,7 +146,7 @@ func (m *Manager) Run(ctx context.Context) {
 
 	m.runOnce.Do(func() {
 		// Closes when Stop() is called or context is canceled.
-		go func() {
+		pproflabel.Go(ctx, pproflabel.Service(pproflabel.ServiceNotifications), func(ctx context.Context) {
 			err := m.loop(ctx)
 			if err != nil {
 				if xerrors.Is(err, ErrManagerAlreadyClosed) {
@@ -154,7 +155,7 @@ func (m *Manager) Run(ctx context.Context) {
 					m.log.Error(ctx, "notification manager stopped with error", slog.Error(err))
 				}
 			}
-		}()
+		})
 	})
 }
 
diff --git a/coderd/pproflabel/pproflabel.go b/coderd/pproflabel/pproflabel.go
@@ -21,9 +21,17 @@ const (
 
 	ServiceHTTPServer           = "http-api"
 	ServiceLifecycles           = "lifecycle-executor"
-	ServiceMetricCollector      = "metrics-collector"
 	ServicePrebuildReconciler   = "prebuilds-reconciler"
 	ServiceTerraformProvisioner = "terraform-provisioner"
+	ServiceDBPurge              = "db-purge"
+	ServiceNotifications        = "notifications"
+	ServiceReplicaSync          = "replica-sync"
+	// ServiceMetricCollector collects metrics from insights in the database and
+	// exports them in a prometheus collector format.
+	ServiceMetricCollector = "metrics-collector"
+	// ServiceAgentMetricAggregator merges agent metrics and exports them in a
+	// prometheus collector format.
+	ServiceAgentMetricAggregator = "agent-metrics-aggregator"
 
 	RequestTypeTag = "coder_request_type"
 )
diff --git a/coderd/prometheusmetrics/aggregator.go b/coderd/prometheusmetrics/aggregator.go
@@ -11,11 +11,11 @@ import (
 	"github.com/prometheus/common/model"
 	"golang.org/x/xerrors"
 
-	"github.com/coder/coder/v2/coderd/agentmetrics"
-
 	"cdr.dev/slog"
 
 	agentproto "github.com/coder/coder/v2/agent/proto"
+	"github.com/coder/coder/v2/coderd/agentmetrics"
+	"github.com/coder/coder/v2/coderd/pproflabel"
 )
 
 const (
@@ -298,7 +298,7 @@ func (ma *MetricsAggregator) Run(ctx context.Context) func() {
 	done := make(chan struct{})
 
 	cleanupTicker := time.NewTicker(ma.metricsCleanupInterval)
-	go func() {
+	pproflabel.Go(ctx, pproflabel.Service(pproflabel.ServiceAgentMetricAggregator), func(ctx context.Context) {
 		defer close(done)
 		defer cleanupTicker.Stop()
 
@@ -395,7 +395,7 @@ func (ma *MetricsAggregator) Run(ctx context.Context) func() {
 				return
 			}
 		}
-	}()
+	})
 	return func() {
 		cancelFunc()
 		<-done
diff --git a/enterprise/replicasync/replicasync.go b/enterprise/replicasync/replicasync.go
@@ -23,6 +23,7 @@ import (
 	"github.com/coder/coder/v2/coderd/database/dbauthz"
 	"github.com/coder/coder/v2/coderd/database/dbtime"
 	"github.com/coder/coder/v2/coderd/database/pubsub"
+	"github.com/coder/coder/v2/coderd/pproflabel"
 )
 
 var PubsubEvent = "replica"
@@ -104,7 +105,7 @@ func New(ctx context.Context, logger slog.Logger, db database.Store, ps pubsub.P
 		return nil, xerrors.Errorf("subscribe: %w", err)
 	}
 	manager.closeWait.Add(1)
-	go manager.loop(ctx)
+	pproflabel.Go(ctx, pproflabel.Service(pproflabel.ServiceReplicaSync), manager.loop)
 	return manager, nil
 }
 

Original file line number	Diff line number	Diff line change
`@@ -23,6 +23,7 @@ import (`
`23`	`23`	`"github.com/coder/coder/v2/coderd/database/dbauthz"`
`24`	`24`	`"github.com/coder/coder/v2/coderd/database/dbtime"`
`25`	`25`	`"github.com/coder/coder/v2/coderd/database/pubsub"`
	`26`	`+ "github.com/coder/coder/v2/coderd/pproflabel"`
`26`	`27`	`)`
`27`	`28`
`28`	`29`	`var PubsubEvent = "replica"`
`@@ -104,7 +105,7 @@ func New(ctx context.Context, logger slog.Logger, db database.Store, ps pubsub.P`
`104`	`105`	`return nil, xerrors.Errorf("subscribe: %w", err)`
`105`	`106`	`}`
`106`	`107`	`manager.closeWait.Add(1)`
`107`		`- go manager.loop(ctx)`
	`108`	`+ pproflabel.Go(ctx, pproflabel.Service(pproflabel.ServiceReplicaSync), manager.loop)`
`108`	`109`	`return manager, nil`
`109`	`110`	`}`
`110`	`111`