chore: add histograms for replication tasks lag, returned, and returned diff

Dual-emit exponential histograms alongside existing timers for
ReplicationTasksLag, ReplicationTasksReturned, and ReplicationTasksReturnedDiff
to support histogram migration. ReplicationTasksLag uses the already-defined
ExponentialReplicationTasksLag (duration-based Mid1ms24h). Returned and
ReturnedDiff use new integer histograms (Mid1To16k) with _counts suffix.
All six metric names added to HistogramMigrationMetrics allowlist.

Signed-off-by: Diana Zawadzki <dzawa@live.de>

Author: zawadzkidiana

common/metrics/config.go

@@ -51,6 +51,15 @@ var HistogramMigrationMetrics = map[string]struct{}{
 
 	"replication_task_latency":    {},
 	"replication_task_latency_ns": {},
+
+	// Replication tasks lag/returned/diff (replication task-ack path).
+	// Dual-emitted as timer + histogram.
+	"replication_tasks_lag":                {},
+	"replication_tasks_lag_ns":             {},
+	"replication_tasks_returned":           {},
+	"replication_tasks_returned_counts":    {},
+	"replication_tasks_returned_diff":      {},
+	"replication_tasks_returned_diff_counts": {},
 }
 
 func (h HistogramMigration) EmitTimer(name string) bool {

common/metrics/defs.go

@@ -2731,7 +2731,9 @@ const (
 	ReplicationTasksDelay
 	ReplicationTasksFetched
 	ReplicationTasksReturned
+	ExponentialReplicationTasksReturned
 	ReplicationTasksReturnedDiff
+	ExponentialReplicationTasksReturnedDiff
 	ReplicationTasksAppliedLatency
 	ExponentialReplicationTasksAppliedLatency
 	ReplicationTasksBatchSize
@@ -3548,7 +3550,9 @@ var MetricDefs = map[ServiceIdx]map[MetricIdx]metricDefinition{
 		ReplicationTasksDelay:                                        {metricName: "replication_tasks_delay", metricType: Histogram, buckets: ReplicationTaskDelayBucket},
 		ReplicationTasksFetched:                                      {metricName: "replication_tasks_fetched", metricType: Timer},
 		ReplicationTasksReturned:                                     {metricName: "replication_tasks_returned", metricType: Timer},
+		ExponentialReplicationTasksReturned:                          {metricName: "replication_tasks_returned_counts", metricType: Histogram, intExponentialBuckets: Mid1To16k},
 		ReplicationTasksReturnedDiff:                                 {metricName: "replication_tasks_returned_diff", metricType: Timer},
+		ExponentialReplicationTasksReturnedDiff:                      {metricName: "replication_tasks_returned_diff_counts", metricType: Histogram, intExponentialBuckets: Mid1To16k},
 		ReplicationTasksAppliedLatency:                               {metricName: "replication_tasks_applied_latency", metricType: Timer},
 		ExponentialReplicationTasksAppliedLatency:                    {metricName: "replication_tasks_applied_latency_ns", metricType: Histogram, exponentialBuckets: Low1ms100s},
 		ReplicationTasksBatchSize:                                    {metricName: "replication_tasks_batch_size", metricType: Gauge},

service/history/replication/task_ack_manager.go

@@ -199,9 +199,18 @@ func (t *TaskAckManager) getTasks(ctx context.Context, pollingCluster string, la
 		return nil, err
 	}
 
-	t.scope.RecordTimer(metrics.ReplicationTasksLag, time.Duration(t.ackLevels.UpdateIfNeededAndGetQueueMaxReadLevel(persistence.HistoryTaskCategoryReplication, pollingCluster).GetTaskID()-msgs.LastRetrievedMessageID))
-	t.scope.RecordTimer(metrics.ReplicationTasksReturned, time.Duration(len(msgs.ReplicationTasks)))
-	t.scope.RecordTimer(metrics.ReplicationTasksReturnedDiff, time.Duration(len(taskInfos)-len(msgs.ReplicationTasks)))
+	// Keep timers (backwards compatible), dual-emit exponential histograms for migration.
+	replicationLag := time.Duration(t.ackLevels.UpdateIfNeededAndGetQueueMaxReadLevel(persistence.HistoryTaskCategoryReplication, pollingCluster).GetTaskID() - msgs.LastRetrievedMessageID)
+	t.scope.RecordTimer(metrics.ReplicationTasksLag, replicationLag)
+	t.scope.ExponentialHistogram(metrics.ExponentialReplicationTasksLag, replicationLag)
+
+	tasksReturned := len(msgs.ReplicationTasks)
+	t.scope.RecordTimer(metrics.ReplicationTasksReturned, time.Duration(tasksReturned))
+	t.scope.IntExponentialHistogram(metrics.ExponentialReplicationTasksReturned, tasksReturned)
+
+	tasksReturnedDiff := len(taskInfos) - len(msgs.ReplicationTasks)
+	t.scope.RecordTimer(metrics.ReplicationTasksReturnedDiff, time.Duration(tasksReturnedDiff))
+	t.scope.IntExponentialHistogram(metrics.ExponentialReplicationTasksReturnedDiff, tasksReturnedDiff)
 
 	t.ackLevel(pollingCluster, lastReadTaskID)