diff --git a/plugins/processors/topk/topk.go b/plugins/processors/topk/topk.go
index 5ba973cba..05fa1ecb5 100644
--- a/plugins/processors/topk/topk.go
+++ b/plugins/processors/topk/topk.go
@@ -163,7 +163,7 @@ func (t *TopK) Apply(in ...telegraf.Metric) []telegraf.Metric {
 		// holding undelivered metrics while the input waits for metrics to be
 		// delivered.  Instead, treat all handled metrics as delivered and
 		// produced metrics as untracked in a similar way to aggregators.
-		m.Drop()
+		m.Accept()
 
 		// Check if the metric has any of the fields over which we are aggregating
 		hasField := false
diff --git a/plugins/processors/topk/topk_test.go b/plugins/processors/topk/topk_test.go
index 8e2550a35..de148a158 100644
--- a/plugins/processors/topk/topk_test.go
+++ b/plugins/processors/topk/topk_test.go
@@ -1,11 +1,15 @@
 package topk
 
 import (
+	"sync"
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/require"
+
 	"github.com/influxdata/telegraf"
 	"github.com/influxdata/telegraf/config"
+	"github.com/influxdata/telegraf/metric"
 	"github.com/influxdata/telegraf/testutil"
 )
 
@@ -55,11 +59,11 @@ func generateAns(input []telegraf.Metric, changeSet map[int]metricChange) []tele
 
 	// For every input metric, we check if there is a change we need to apply
 	// If there is no change for a given input metric, the metric is dropped
-	for i, metric := range input {
+	for i, m := range input {
 		change, ok := changeSet[i]
 		if ok {
 			// Deep copy the metric
-			newMetric := metric.Copy()
+			newMetric := m.Copy()
 
 			// Add new fields
 			if change.newFields != nil {
@@ -501,3 +505,77 @@ func TestTopkGroupByKeyTag(t *testing.T) {
 	// Run the test
 	runAndCompare(&topk, input, answer, "GroupByKeyTag test", t)
 }
+
+func TestTracking(t *testing.T) {
+	inputRaw := []telegraf.Metric{
+		metric.New("foo", map[string]string{}, map[string]interface{}{"value": 100}, time.Unix(0, 0)),
+		metric.New("bar", map[string]string{}, map[string]interface{}{"value": 22}, time.Unix(0, 0)),
+		metric.New("baz", map[string]string{}, map[string]interface{}{"value": 1}, time.Unix(0, 0)),
+	}
+
+	var mu sync.Mutex
+	delivered := make([]telegraf.DeliveryInfo, 0, len(inputRaw))
+	notify := func(di telegraf.DeliveryInfo) {
+		mu.Lock()
+		defer mu.Unlock()
+		delivered = append(delivered, di)
+	}
+
+	input := make([]telegraf.Metric, 0, len(inputRaw))
+	for _, m := range inputRaw {
+		tm, _ := metric.WithTracking(m, notify)
+		input = append(input, tm)
+	}
+
+	expected := []telegraf.Metric{
+		metric.New(
+			"foo",
+			map[string]string{},
+			map[string]interface{}{"value": 100},
+			time.Unix(0, 0),
+		),
+		metric.New(
+			"bar",
+			map[string]string{},
+			map[string]interface{}{"value": 22},
+			time.Unix(0, 0),
+		),
+		metric.New(
+			"baz",
+			map[string]string{},
+			map[string]interface{}{"value": 1},
+			time.Unix(0, 0),
+		),
+	}
+
+	// Only doing this over 1 period, so we should expect the same number of
+	// metrics back.
+	plugin := &TopK{
+		Period:      1,
+		K:           3,
+		Aggregation: "mean",
+		Fields:      []string{"value"},
+		Log:         testutil.Logger{},
+	}
+	plugin.Reset()
+
+	// Process expected metrics and compare with resulting metrics
+	var actual []telegraf.Metric
+	require.Eventuallyf(t, func() bool {
+		actual = plugin.Apply(input...)
+		return len(actual) > 0
+	}, time.Second, 100*time.Millisecond, "never got any metrics")
+	testutil.RequireMetricsEqual(t, expected, actual)
+
+	// Simulate output acknowledging delivery
+	for _, m := range actual {
+		m.Accept()
+	}
+
+	// Check delivery
+	require.Eventuallyf(t, func() bool {
+		mu.Lock()
+		defer mu.Unlock()
+		return len(expected) == len(delivered)
+	}, time.Second, 100*time.Millisecond, "%d delivered but %d expected", len(delivered), len(expected))
+}