Use generics in jitter

common/backoff/jitter.go

@@ -26,39 +26,26 @@ package backoff
 
 import (
 	"math/rand"
-	"time"
 )
 
-// JitDuration return random duration from (1-coefficient)*duration to (1+coefficient)*duration, inclusive, exclusive
-func JitDuration(duration time.Duration, coefficient float64) time.Duration {
-	validateCoefficient(coefficient)
+const fullCoefficient float64 = 1
 
-	return time.Duration(JitInt64(duration.Nanoseconds(), coefficient))
+// FullJitter return random number from 0 to input, inclusive, exclusive
+func FullJitter[T ~int64 | ~int | ~int32 | ~float64 | ~float32](input T) T {
+	return Jitter(input, fullCoefficient) / 2
 }
 
-// JitInt64 return random number from (1-coefficient)*input to (1+coefficient)*input, inclusive, exclusive
-func JitInt64(input int64, coefficient float64) int64 {
+// Jitter return random number from (1-coefficient)*input to (1+coefficient)*input, inclusive, exclusive
+func Jitter[T ~int64 | ~int | ~int32 | ~float64 | ~float32](input T, coefficient float64) T {
 	validateCoefficient(coefficient)
 
-	if input == 0 {
-		return 0
-	}
 	if coefficient == 0 {
 		return input
 	}
 
-	base := int64(float64(input) * (1 - coefficient))
-	addon := rand.Int63n(2 * (input - base))
-	return base + addon
-}
-
-// JitFloat64 return random number from (1-coefficient)*input to (1+coefficient)*input, inclusive, exclusive
-func JitFloat64(input float64, coefficient float64) float64 {
-	validateCoefficient(coefficient)
-
-	base := input * (1 - coefficient)
-	addon := rand.Float64() * 2 * (input - base)
-	return base + addon
+	base := float64(input) * (1 - coefficient)
+	addon := rand.Float64() * 2 * (float64(input) - base)
+	return T(base + addon)
 }
 
 func validateCoefficient(coefficient float64) {

common/backoff/jitter_test.go

@@ -46,47 +46,68 @@ func TestJitterSuite(t *testing.T) {
 func (s *jitterSuite) SetupSuite() {
 }
 
-func (s *jitterSuite) TestJitInt64() {
+func (s *jitterSuite) TestJitter_Int64() {
 	input := int64(1048576)
 	coefficient := float64(0.25)
 	lowerBound := int64(float64(input) * (1 - coefficient))
 	upperBound := int64(float64(input) * (1 + coefficient))
+	fullJitterUpperBound := int64(float64(input) * 2)
 
 	for i := 0; i < 1048576; i++ {
-		result := JitInt64(input, coefficient)
+		result := Jitter(input, coefficient)
 		s.True(result >= lowerBound)
 		s.True(result < upperBound)
+
+		result = FullJitter(input)
+		s.True(result >= 0)
+		s.True(result < fullJitterUpperBound)
 	}
 }
 
-func (s *jitterSuite) TestJitFloat64() {
+func (s *jitterSuite) TestJitter_Float64() {
 	input := float64(1048576.1048576)
 	coefficient := float64(0.16)
 	lowerBound := float64(input) * (1 - coefficient)
 	upperBound := float64(input) * (1 + coefficient)
+	fullJitterUpperBound := float64(input) * 2
 
 	for i := 0; i < 1048576; i++ {
-		result := JitFloat64(input, coefficient)
+		result := Jitter(input, coefficient)
 		s.True(result >= lowerBound)
 		s.True(result < upperBound)
+
+		result = FullJitter(input)
+		s.True(result >= 0)
+		s.True(result < fullJitterUpperBound)
 	}
 }
 
-func (s *jitterSuite) TestJitDuration() {
+func (s *jitterSuite) TestJitter_Duration() {
 	input := time.Duration(1099511627776)
 	coefficient := float64(0.1)
 	lowerBound := time.Duration(int64(float64(input.Nanoseconds()) * (1 - coefficient)))
 	upperBound := time.Duration(int64(float64(input.Nanoseconds()) * (1 + coefficient)))
+	fullJitterUpperBound := time.Duration(int64(float64(input.Nanoseconds()) * 2))
 
 	for i := 0; i < 1048576; i++ {
-		result := JitDuration(input, coefficient)
+		result := Jitter(input, coefficient)
 		s.True(result >= lowerBound)
 		s.True(result < upperBound)
+
+		result = FullJitter(input)
+		s.True(result >= 0)
+		s.True(result < fullJitterUpperBound)
 	}
 }
 
-func (s *jitterSuite) TestJit_InputZeroValue() {
-	s.Zero(JitDuration(0, rand.Float64()))
-	s.Zero(JitInt64(0, rand.Float64()))
-	s.Zero(JitFloat64(0, rand.Float64()))
+func (s *jitterSuite) TestJitter_InputZeroValue() {
+	s.Zero(Jitter(time.Duration(0), rand.Float64()))
+	s.Zero(Jitter(int64(0), rand.Float64()))
+	s.Zero(Jitter(float64(0), rand.Float64()))
+}
+
+func (s *jitterSuite) TestJitter_CoeffientZeroValue() {
+	s.Equal(time.Duration(1), Jitter(time.Duration(1), 0))
+	s.Equal(int64(1), Jitter(int64(1), 0))
+	s.Equal(float64(1), Jitter(float64(1), 0))
 }

common/tasks/fifo_scheduler.go

@@ -145,7 +145,7 @@ func (f *FIFOScheduler[T]) TrySubmit(task T) bool {
 func (f *FIFOScheduler[T]) workerMonitor() {
 	defer f.shutdownWG.Done()
 
-	timer := time.NewTimer(backoff.JitDuration(defaultMonitorTickerDuration, defaultMonitorTickerJitter))
+	timer := time.NewTimer(backoff.Jitter(defaultMonitorTickerDuration, defaultMonitorTickerJitter))
 	defer timer.Stop()
 
 	for {
@@ -154,7 +154,7 @@ func (f *FIFOScheduler[T]) workerMonitor() {
 			f.stopWorkers(len(f.workerShutdownCh))
 			return
 		case <-timer.C:
-			timer.Reset(backoff.JitDuration(defaultMonitorTickerDuration, defaultMonitorTickerJitter))
+			timer.Reset(backoff.Jitter(defaultMonitorTickerDuration, defaultMonitorTickerJitter))
 
 			targetWorkerNum := f.options.WorkerCount()
 			currentWorkerNum := len(f.workerShutdownCh)

common/tasks/interleaved_weighted_round_robin.go

@@ -312,7 +312,7 @@ func (s *InterleavedWeightedRoundRobinScheduler[T, K]) channels() iwrrChannels[T
 
 func (s *InterleavedWeightedRoundRobinScheduler[T, K]) setupDispatchTimer() {
 	throttleDuration := iwrrMinDispatchThrottleDuration +
-		backoff.JitDuration(s.options.MaxDispatchThrottleDuration-iwrrMinDispatchThrottleDuration, 1)/2
+		backoff.FullJitter(s.options.MaxDispatchThrottleDuration-iwrrMinDispatchThrottleDuration)
 
 	s.dispatchTimerLock.Lock()
 	defer s.dispatchTimerLock.Unlock()

service/history/queues/queue_base.go

@@ -265,7 +265,7 @@ func (p *queueBase) Start() {
 	p.rescheduler.Start()
 	p.readerGroup.Start()
 
-	p.checkpointTimer = time.NewTimer(backoff.JitDuration(
+	p.checkpointTimer = time.NewTimer(backoff.Jitter(
 		p.options.CheckpointInterval(),
 		p.options.CheckpointIntervalJitterCoefficient(),
 	))
@@ -442,7 +442,7 @@ func (p *queueBase) resetCheckpointTimer(checkPointErr error) {
 	}
 
 	p.checkpointRetrier.Reset()
-	p.checkpointTimer.Reset(backoff.JitDuration(
+	p.checkpointTimer.Reset(backoff.Jitter(
 		p.options.CheckpointInterval(),
 		p.options.CheckpointIntervalJitterCoefficient(),
 	))

service/history/queues/queue_immediate.go

@@ -146,7 +146,7 @@ func (p *immediateQueue) NotifyNewTasks(_ string, tasks []tasks.Task) {
 func (p *immediateQueue) processEventLoop() {
 	defer p.shutdownWG.Done()
 
-	pollTimer := time.NewTimer(backoff.JitDuration(
+	pollTimer := time.NewTimer(backoff.Jitter(
 		p.options.MaxPollInterval(),
 		p.options.MaxPollIntervalJitterCoefficient(),
 	))
@@ -171,7 +171,7 @@ func (p *immediateQueue) processEventLoop() {
 func (p *immediateQueue) processPollTimer(pollTimer *time.Timer) {
 	p.processNewRange()
 
-	pollTimer.Reset(backoff.JitDuration(
+	pollTimer.Reset(backoff.Jitter(
 		p.options.MaxPollInterval(),
 		p.options.MaxPollIntervalJitterCoefficient(),
 	))

service/history/queues/queue_scheduled.go

@@ -226,7 +226,7 @@ func (p *scheduledQueue) processNewRange() {
 		// in which case no look ahead is needed.
 		// Notification will be sent when shard is reacquired, but
 		// still set a max poll timer here as a catch all case.
-		p.timerGate.Update(p.timeSource.Now().Add(backoff.JitDuration(
+		p.timerGate.Update(p.timeSource.Now().Add(backoff.Jitter(
 			p.options.MaxPollInterval(),
 			p.options.MaxPollIntervalJitterCoefficient(),
 		)))
@@ -251,7 +251,7 @@ func (p *scheduledQueue) lookAheadTask() {
 	}
 
 	lookAheadMinTime := p.nonReadableScope.Range.InclusiveMin.FireTime
-	lookAheadMaxTime := lookAheadMinTime.Add(backoff.JitDuration(
+	lookAheadMaxTime := lookAheadMinTime.Add(backoff.Jitter(
 		p.options.MaxPollInterval(),
 		p.options.MaxPollIntervalJitterCoefficient(),
 	))

service/history/queues/rescheduler.go

@@ -198,7 +198,7 @@ func (r *reschedulerImpl) Len() int {
 func (r *reschedulerImpl) rescheduleLoop() {
 	defer r.shutdownWG.Done()
 
-	cleanupTimer := time.NewTimer(backoff.JitDuration(
+	cleanupTimer := time.NewTimer(backoff.Jitter(
 		reschedulerPQCleanupDuration,
 		reschedulerPQCleanupJitterCoefficient,
 	))
@@ -213,7 +213,7 @@ func (r *reschedulerImpl) rescheduleLoop() {
 			r.reschedule()
 		case <-cleanupTimer.C:
 			r.cleanupPQ()
-			cleanupTimer.Reset(backoff.JitDuration(
+			cleanupTimer.Reset(backoff.Jitter(
 				reschedulerPQCleanupDuration,
 				reschedulerPQCleanupJitterCoefficient,
 			))

service/history/replication/ack_manager.go

@@ -364,7 +364,7 @@ func (p *ackMgrImpl) taskIDsRange(
 
 	now := p.shard.GetTimeSource().Now()
 	if p.sanityCheckTime.IsZero() || p.sanityCheckTime.Before(now) {
-		p.sanityCheckTime = now.Add(backoff.JitDuration(
+		p.sanityCheckTime = now.Add(backoff.Jitter(
 			p.config.ReplicatorProcessorMaxPollInterval(),
 			p.config.ReplicatorProcessorMaxPollIntervalJitterCoefficient(),
 		))

service/history/replication/task_fetcher.go

@@ -31,6 +31,8 @@ import (
 	"sync/atomic"
 	"time"
 
+	"golang.org/x/exp/maps"
+
 	"go.temporal.io/server/api/adminservice/v1"
 	replicationspb "go.temporal.io/server/api/replication/v1"
 	"go.temporal.io/server/client"
@@ -43,7 +45,6 @@ import (
 	"go.temporal.io/server/common/quotas"
 	"go.temporal.io/server/common/rpc"
 	"go.temporal.io/server/service/history/configs"
-	"golang.org/x/exp/maps"
 )
 
 const (
@@ -356,7 +357,7 @@ func (f *replicationTaskFetcherWorker) Stop() {
 
 // fetchTasks collects getReplicationTasks request from shards and send out aggregated request to source frontend.
 func (f *replicationTaskFetcherWorker) fetchTasks() {
-	timer := time.NewTimer(backoff.JitDuration(
+	timer := time.NewTimer(backoff.Jitter(
 		f.config.ReplicationTaskFetcherAggregationInterval(),
 		f.config.ReplicationTaskFetcherTimerJitterCoefficient(),
 	))
@@ -371,12 +372,12 @@ func (f *replicationTaskFetcherWorker) fetchTasks() {
 			// When timer fires, we collect all the requests we have so far and attempt to send them to remote.
 			err := f.getMessages()
 			if err != nil {
-				timer.Reset(backoff.JitDuration(
+				timer.Reset(backoff.Jitter(
 					f.config.ReplicationTaskFetcherErrorRetryWait(),
 					f.config.ReplicationTaskFetcherTimerJitterCoefficient(),
 				))
 			} else {
-				timer.Reset(backoff.JitDuration(
+				timer.Reset(backoff.Jitter(
 					f.config.ReplicationTaskFetcherAggregationInterval(),
 					f.config.ReplicationTaskFetcherTimerJitterCoefficient(),
 				))

service/history/replication/task_processor.go

@@ -189,7 +189,7 @@ func (p *taskProcessorImpl) Stop() {
 }
 
 func (p *taskProcessorImpl) eventLoop() {
-	syncShardTimer := time.NewTimer(backoff.JitDuration(
+	syncShardTimer := time.NewTimer(backoff.Jitter(
 		p.config.ShardSyncMinInterval(),
 		p.config.ShardSyncTimerJitterCoefficient(),
 	))
@@ -210,7 +210,7 @@ func (p *taskProcessorImpl) eventLoop() {
 					1,
 					metrics.OperationTag(metrics.HistorySyncShardStatusScope))
 			}
-			syncShardTimer.Reset(backoff.JitDuration(
+			syncShardTimer.Reset(backoff.Jitter(
 				p.config.ShardSyncMinInterval(),
 				p.config.ShardSyncTimerJitterCoefficient(),
 			))

service/history/replication/task_processor_manager.go

@@ -204,7 +204,7 @@ func (r *taskProcessorManagerImpl) handleClusterMetadataUpdate(
 
 func (r *taskProcessorManagerImpl) completeReplicationTaskLoop() {
 	shardID := r.shard.GetShardID()
-	cleanupTimer := time.NewTimer(backoff.JitDuration(
+	cleanupTimer := time.NewTimer(backoff.Jitter(
 		r.config.ReplicationTaskProcessorCleanupInterval(shardID),
 		r.config.ReplicationTaskProcessorCleanupJitterCoefficient(shardID),
 	))
@@ -219,7 +219,7 @@ func (r *taskProcessorManagerImpl) completeReplicationTaskLoop() {
 					metrics.OperationTag(metrics.ReplicationTaskCleanupScope),
 				)
 			}
-			cleanupTimer.Reset(backoff.JitDuration(
+			cleanupTimer.Reset(backoff.Jitter(
 				r.config.ReplicationTaskProcessorCleanupInterval(shardID),
 				r.config.ReplicationTaskProcessorCleanupJitterCoefficient(shardID),
 			))

service/history/workflow/task_generator.go

@@ -148,10 +148,6 @@ func (r *TaskGeneratorImpl) GenerateWorkflowStartTasks(
 	return nil
 }
 
-// archivalDelayJitterCoefficient is a variable because we need to override it to 0 in unit tests to make them
-// deterministic.
-var archivalDelayJitterCoefficient = 1.0
-
 func (r *TaskGeneratorImpl) GenerateWorkflowCloseTasks(
 	closeEvent *historypb.HistoryEvent,
 	deleteAfterClose bool,
@@ -198,7 +194,8 @@ func (r *TaskGeneratorImpl) GenerateWorkflowCloseTasks(
 			}
 			// We schedule the archival task for a random time in the near future to avoid sending a surge of tasks
 			// to the archival system at the same time
-			delay := backoff.JitDuration(r.config.ArchivalProcessorArchiveDelay(), archivalDelayJitterCoefficient) / 2
+
+			delay := backoff.FullJitter(r.config.ArchivalProcessorArchiveDelay())
 			if delay > retention {
 				delay = retention
 			}
@@ -258,7 +255,7 @@ func (r *TaskGeneratorImpl) GenerateDeleteHistoryEventTask(closeTime time.Time,
 		return err
 	}
 
-	retentionJitterDuration := backoff.JitDuration(r.config.RetentionTimerJitterDuration(), 1) / 2
+	retentionJitterDuration := backoff.FullJitter(r.config.RetentionTimerJitterDuration())
 	deleteTime := closeTime.Add(retention).Add(retentionJitterDuration)
 	r.mutableState.AddTasks(&tasks.DeleteHistoryEventTask{
 		// TaskID is set by shard

service/history/workflow/task_generator_test.go

@@ -87,8 +87,6 @@ type testParams struct {
 }
 
 func TestTaskGeneratorImpl_GenerateWorkflowCloseTasks(t *testing.T) {
-	// we need to set the jitter coefficient to 0 to remove the randomness in the test
-	archivalDelayJitterCoefficient = 0.0
 	for _, c := range []testConfig{
 		{
 			Name: "delete after retention",
@@ -253,11 +251,8 @@ func TestTaskGeneratorImpl_GenerateWorkflowCloseTasks(t *testing.T) {
 				assert.Equal(t, archiveExecutionTask.NamespaceID, namespaceEntry.ID().String())
 				assert.Equal(t, archiveExecutionTask.WorkflowID, tests.WorkflowID)
 				assert.Equal(t, archiveExecutionTask.RunID, tests.RunID)
-				assert.Equal(
-					t,
-					p.ExpectedArchiveExecutionTaskVisibilityTimestamp,
-					archiveExecutionTask.VisibilityTimestamp,
-				)
+				assert.True(t, p.ExpectedArchiveExecutionTaskVisibilityTimestamp.Equal(archiveExecutionTask.VisibilityTimestamp) ||
+					p.ExpectedArchiveExecutionTaskVisibilityTimestamp.After(archiveExecutionTask.VisibilityTimestamp))
 			} else {
 				assert.Nil(t, archiveExecutionTask)
 			}

service/worker/replicator/namespace_replication_message_processor.go

@@ -225,7 +225,7 @@ func (p *namespaceReplicationMessageProcessor) Stop() {
 }
 
 func getWaitDuration() time.Duration {
-	return backoff.JitDuration(time.Duration(pollIntervalSecs)*time.Second, pollTimerJitterCoefficient)
+	return backoff.Jitter(time.Duration(pollIntervalSecs)*time.Second, pollTimerJitterCoefficient)
 }
 
 func isTransientRetryableError(err error) bool {

service/worker/scanner/executions/task.go

@@ -106,7 +106,7 @@ func newTask(
 
 // Run runs the task
 func (t *task) Run() executor.TaskStatus {
-	time.Sleep(backoff.JitDuration(
+	time.Sleep(backoff.Jitter(
 		taskStartupDelayRatio*time.Duration(t.scavenger.numHistoryShards),
 		taskStartupDelayRandomizationRatio,
 	))