cleanup: prevent leaks from time.After

This PR replaces use of time.After with a safe helper function
that creates a time.Timer to use instead. The new function returns
both a time.Timer and a Stop function that the caller must handle.

Unlike time.NewTimer, the helper function does not panic if the duration
set is <= 0.

api/allocations_exec.go

@@ -13,6 +13,12 @@ import (
 	"github.com/gorilla/websocket"
 )
 
+const (
+	// heartbeatInterval is the amount of time to wait between sending heartbeats
+	// during an exec streaming operation
+	heartbeatInterval = 10 * time.Second
+)
+
 type execSession struct {
 	client  *Client
 	alloc   *Allocation
@@ -177,15 +183,19 @@ func (s *execSession) startTransmit(ctx context.Context, conn *websocket.Conn) <
 
 	// send a heartbeat every 10 seconds
 	go func() {
+		t := time.NewTimer(heartbeatInterval)
+		defer t.Stop()
+
 		for {
+			t.Reset(heartbeatInterval)
+
 			select {
 			case <-ctx.Done():
 				return
-			// heartbeat message
-			case <-time.After(10 * time.Second):
+			case <-t.C:
+				// heartbeat message
 				send(&execStreamingInputHeartbeat)
 			}
-
 		}
 	}()
 

client/allocrunner/taskrunner/task_runner.go

@@ -527,6 +527,9 @@ func (tr *TaskRunner) Run() {
 		return
 	}
 
+	timer, stop := helper.NewSafeTimer(0) // timer duration calculated JIT
+	defer stop()
+
 MAIN:
 	for !tr.shouldShutdown() {
 		select {
@@ -612,9 +615,11 @@ MAIN:
 			break MAIN
 		}
 
+		timer.Reset(restartDelay)
+
 		// Actually restart by sleeping and also watching for destroy events
 		select {
-		case <-time.After(restartDelay):
+		case <-timer.C:
 		case <-tr.killCtx.Done():
 			tr.logger.Trace("task killed between restarts", "delay", restartDelay)
 			break MAIN

command/agent/consul/version_checker.go

@@ -7,6 +7,7 @@ import (
 
 	log "github.com/hashicorp/go-hclog"
 	version "github.com/hashicorp/go-version"
+	"github.com/hashicorp/nomad/helper"
 )
 
 // checkConsulTLSSkipVerify logs if Consul does not support TLSSkipVerify on
@@ -20,6 +21,10 @@ func checkConsulTLSSkipVerify(ctx context.Context, logger log.Logger, client Age
 	defer close(done)
 
 	i := uint64(0)
+
+	timer, stop := helper.NewSafeTimer(limit)
+	defer stop()
+
 	for {
 		self, err := client.Self()
 		if err == nil {
@@ -39,10 +44,12 @@ func checkConsulTLSSkipVerify(ctx context.Context, logger log.Logger, client Age
 			i++
 		}
 
+		timer.Reset(backoff)
+
 		select {
 		case <-ctx.Done():
 			return
-		case <-time.After(backoff):
+		case <-timer.C:
 		}
 	}
 }

command/agent/monitor/monitor.go

@@ -6,6 +6,7 @@ import (
 	"time"
 
 	log "github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/nomad/helper"
 )
 
 // Monitor provides a mechanism to stream logs using go-hclog
@@ -107,12 +108,17 @@ func (d *monitor) Start() <-chan []byte {
 	// dropped messages and makes room on the logCh
 	// to add a dropped message count warning
 	go func() {
+		timer, stop := helper.NewSafeTimer(d.droppedDuration)
+		defer stop()
+
 		// loop and check for dropped messages
 		for {
+			timer.Reset(d.droppedDuration)
+
 			select {
 			case <-d.doneCh:
 				return
-			case <-time.After(d.droppedDuration):
+			case <-timer.C:
 				d.Lock()
 
 				// Check if there have been any dropped messages.

drivers/docker/stats.go

@@ -10,6 +10,7 @@ import (
 	docker "github.com/fsouza/go-dockerclient"
 	cstructs "github.com/hashicorp/nomad/client/structs"
 	"github.com/hashicorp/nomad/drivers/docker/util"
+	"github.com/hashicorp/nomad/helper"
 	nstructs "github.com/hashicorp/nomad/nomad/structs"
 )
 
@@ -93,17 +94,25 @@ func (h *taskHandle) collectStats(ctx context.Context, destCh *usageSender, inte
 	// backoff and retry used if the docker stats API returns an error
 	var backoff time.Duration
 	var retry int
+
+	// create an interval timer
+	timer, stop := helper.NewSafeTimer(interval)
+	defer stop()
+
 	// loops until doneCh is closed
 	for {
+		timer.Reset(interval)
+
 		if backoff > 0 {
 			select {
-			case <-time.After(backoff):
+			case <-timer.C:
 			case <-ctx.Done():
 				return
 			case <-h.doneCh:
 				return
 			}
 		}
+
 		// make a channel for docker stats structs and start a collector to
 		// receive stats from docker and emit nomad stats
 		// statsCh will always be closed by docker client.

drivers/shared/eventer/eventer.go

@@ -6,6 +6,7 @@ import (
 	"time"
 
 	hclog "github.com/hashicorp/go-hclog"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/plugins/drivers"
 )
 
@@ -62,14 +63,19 @@ func NewEventer(ctx context.Context, logger hclog.Logger) *Eventer {
 // eventLoop is the main logic which pulls events from the channel and broadcasts
 // them to all consumers
 func (e *Eventer) eventLoop() {
+	timer, stop := helper.NewSafeTimer(ConsumerGCInterval)
+	defer stop()
+
 	for {
+		timer.Reset(ConsumerGCInterval)
+
 		select {
 		case <-e.ctx.Done():
 			e.logger.Trace("task event loop shutdown")
 			return
 		case event := <-e.events:
 			e.iterateConsumers(event)
-		case <-time.After(ConsumerGCInterval):
+		case <-timer.C:
 			e.gcConsumers()
 		}
 	}

helper/funcs.go

@@ -572,3 +572,30 @@ func PathEscapesSandbox(sandboxDir, path string) bool {
 	}
 	return false
 }
+
+// TimerStopFunc is used to stop a time.Timer created with NewSafeTimer
+type TimerStopFunc func()
+
+// NewSafeTimer creates a time.Timer but does not panic if duration is <= 0.
+//
+// Using a time.Timer is recommended instead of time.After when it is necessary
+// to avoid leaking goroutines (e.g. in a select inside a loop).
+//
+// Returns the time.Timer and also a TimerStopFunc, forcing the caller to deal
+// with stopping the time.Timer to avoid leaking a goroutine.
+func NewSafeTimer(duration time.Duration) (*time.Timer, TimerStopFunc) {
+	if duration <= 0 {
+		// Avoid panic by using the smallest positive value. This is close enough
+		// to the behavior of time.After(0), which this helper is intended to
+		// replace.
+		// https://go.dev/play/p/EIkm9MsPbHY
+		duration = 1
+	}
+
+	t := time.NewTimer(duration)
+	cancel := func() {
+		t.Stop()
+	}
+
+	return t, cancel
+}

helper/funcs_test.go

@@ -431,3 +431,17 @@ func TestPathEscapesSandbox(t *testing.T) {
 		})
 	}
 }
+
+func Test_NewSafeTimer(t *testing.T) {
+	t.Run("zero", func(t *testing.T) {
+		timer, stop := NewSafeTimer(0)
+		defer stop()
+		<-timer.C
+	})
+
+	t.Run("positive", func(t *testing.T) {
+		timer, stop := NewSafeTimer(1)
+		defer stop()
+		<-timer.C
+	})
+}

nomad/blocked_evals.go

@@ -706,9 +706,14 @@ func (b *BlockedEvals) Stats() *BlockedStats {
 
 // EmitStats is used to export metrics about the blocked eval tracker while enabled
 func (b *BlockedEvals) EmitStats(period time.Duration, stopCh <-chan struct{}) {
+	timer, stop := helper.NewSafeTimer(period)
+	defer stop()
+
 	for {
+		timer.Reset(period)
+
 		select {
-		case <-time.After(period):
+		case <-timer.C:
 			stats := b.Stats()
 			metrics.SetGauge([]string{"nomad", "blocked_evals", "total_quota_limit"}, float32(stats.TotalQuotaLimit))
 			metrics.SetGauge([]string{"nomad", "blocked_evals", "total_blocked"}, float32(stats.TotalBlocked))

nomad/drainer/watch_jobs.go

@@ -140,10 +140,16 @@ func (w *drainingJobWatcher) deregisterJob(jobID, namespace string) {
 
 // watch is the long lived watching routine that detects job drain changes.
 func (w *drainingJobWatcher) watch() {
+	timer, stop := helper.NewSafeTimer(stateReadErrorDelay)
+	defer stop()
+
 	waitIndex := uint64(1)
+
 	for {
+		timer.Reset(stateReadErrorDelay)
 		w.logger.Trace("getting job allocs at index", "index", waitIndex)
 		jobAllocs, index, err := w.getJobAllocs(w.getQueryCtx(), waitIndex)
+
 		if err != nil {
 			if err == context.Canceled {
 				// Determine if it is a cancel or a shutdown

nomad/drainer/watch_nodes.go

@@ -2,10 +2,10 @@ package drainer
 
 import (
 	"context"
-	"time"
 
 	log "github.com/hashicorp/go-hclog"
 	memdb "github.com/hashicorp/go-memdb"
+	"github.com/hashicorp/nomad/helper"
 
 	"github.com/hashicorp/nomad/nomad/state"
 	"github.com/hashicorp/nomad/nomad/structs"
@@ -148,8 +148,13 @@ func NewNodeDrainWatcher(ctx context.Context, limiter *rate.Limiter, state *stat
 
 // watch is the long lived watching routine that detects node changes.
 func (w *nodeDrainWatcher) watch() {
+	timer, stop := helper.NewSafeTimer(stateReadErrorDelay)
+	defer stop()
+
 	nindex := uint64(1)
+
 	for {
+		timer.Reset(stateReadErrorDelay)
 		nodes, index, err := w.getNodes(nindex)
 		if err != nil {
 			if err == context.Canceled {
@@ -160,7 +165,7 @@ func (w *nodeDrainWatcher) watch() {
 			select {
 			case <-w.ctx.Done():
 				return
-			case <-time.After(stateReadErrorDelay):
+			case <-timer.C:
 				continue
 			}
 		}

nomad/eval_broker.go

@@ -2,15 +2,15 @@ package nomad
 
 import (
 	"container/heap"
+	"context"
 	"errors"
 	"fmt"
 	"math/rand"
 	"sync"
 	"time"
 
-	"context"
-
 	metrics "github.com/armon/go-metrics"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/helper/uuid"
 	"github.com/hashicorp/nomad/lib/delayheap"
 	"github.com/hashicorp/nomad/nomad/structs"
@@ -835,9 +835,14 @@ func (b *EvalBroker) Stats() *BrokerStats {
 
 // EmitStats is used to export metrics about the broker while enabled
 func (b *EvalBroker) EmitStats(period time.Duration, stopCh <-chan struct{}) {
+	timer, stop := helper.NewSafeTimer(period)
+	defer stop()
+
 	for {
+		timer.Reset(period)
+
 		select {
-		case <-time.After(period):
+		case <-timer.C:
 			stats := b.Stats()
 			metrics.SetGauge([]string{"nomad", "broker", "total_ready"}, float32(stats.TotalReady))
 			metrics.SetGauge([]string{"nomad", "broker", "total_unacked"}, float32(stats.TotalUnacked))

nomad/plan_queue.go

@@ -7,6 +7,7 @@ import (
 	"time"
 
 	metrics "github.com/armon/go-metrics"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/nomad/structs"
 )
 
@@ -196,12 +197,14 @@ func (q *PlanQueue) Stats() *QueueStats {
 
 // EmitStats is used to export metrics about the broker while enabled
 func (q *PlanQueue) EmitStats(period time.Duration, stopCh <-chan struct{}) {
+	timer, stop := helper.NewSafeTimer(period)
+	defer stop()
+
 	for {
 		select {
-		case <-time.After(period):
+		case <-timer.C:
 			stats := q.Stats()
 			metrics.SetGauge([]string{"nomad", "plan", "queue_depth"}, float32(stats.Depth))
-
 		case <-stopCh:
 			return
 		}

nomad/server.go

@@ -27,6 +27,7 @@ import (
 	multierror "github.com/hashicorp/go-multierror"
 	lru "github.com/hashicorp/golang-lru"
 	"github.com/hashicorp/nomad/command/agent/consul"
+	"github.com/hashicorp/nomad/helper"
 	"github.com/hashicorp/nomad/helper/codec"
 	"github.com/hashicorp/nomad/helper/pool"
 	"github.com/hashicorp/nomad/helper/stats"
@@ -1852,9 +1853,14 @@ func (s *Server) Stats() map[string]map[string]string {
 
 // EmitRaftStats is used to export metrics about raft indexes and state store snapshot index
 func (s *Server) EmitRaftStats(period time.Duration, stopCh <-chan struct{}) {
+	timer, stop := helper.NewSafeTimer(period)
+	defer stop()
+
 	for {
+		timer.Reset(period)
+
 		select {
-		case <-time.After(period):
+		case <-timer.C:
 			lastIndex := s.raft.LastIndex()
 			metrics.SetGauge([]string{"raft", "lastIndex"}, float32(lastIndex))
 			appliedIndex := s.raft.AppliedIndex()