diff --git a/CHANGELOG.md b/CHANGELOG.md
index ff7f1f31f3..b716c44827 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -11,6 +11,7 @@
 - [#4198](https://github.com/influxdb/influxdb/pull/4198): Add basic cluster-service stats
 - [#4262](https://github.com/influxdb/influxdb/pull/4262): Allow configuration of UDP retention policy
 - [#4265](https://github.com/influxdb/influxdb/pull/4265): Add statistics for Hinted-Handoff
+- [#4284](https://github.com/influxdb/influxdb/pull/4284): Add exponential backoff for hinted-handoff failures
 
 ### Bugfixes
 - [#4166](https://github.com/influxdb/influxdb/pull/4166): Fix parser error on invalid SHOW
diff --git a/services/hh/config.go b/services/hh/config.go
index dfdca1a37e..b5ffe715fe 100644
--- a/services/hh/config.go
+++ b/services/hh/config.go
@@ -19,26 +19,34 @@ const (
 	// value of 0 disables the rate limit.
 	DefaultRetryRateLimit = 0
 
-	// DefaultRetryInterval is the default amout of time the system waits before
-	// attempting to flush hinted handoff queues.
+	// DefaultRetryInterval is the default amount of time the system waits before
+	// attempting to flush hinted handoff queues. With each failure of a hinted
+	// handoff write, this retry interval increases exponentially until it reaches
+	// the maximum
 	DefaultRetryInterval = time.Second
+
+	// DefaultRetryMaxInterval is the maximum the hinted handoff retry interval
+	// will ever be.
+	DefaultRetryMaxInterval = time.Minute
 )
 
 type Config struct {
-	Enabled        bool          `toml:"enabled"`
-	Dir            string        `toml:"dir"`
-	MaxSize        int64         `toml:"max-size"`
-	MaxAge         toml.Duration `toml:"max-age"`
-	RetryRateLimit int64         `toml:"retry-rate-limit"`
-	RetryInterval  toml.Duration `toml:"retry-interval"`
+	Enabled          bool          `toml:"enabled"`
+	Dir              string        `toml:"dir"`
+	MaxSize          int64         `toml:"max-size"`
+	MaxAge           toml.Duration `toml:"max-age"`
+	RetryRateLimit   int64         `toml:"retry-rate-limit"`
+	RetryInterval    toml.Duration `toml:"retry-interval"`
+	RetryMaxInterval toml.Duration `toml:"retry-max-interval"`
 }
 
 func NewConfig() Config {
 	return Config{
-		Enabled:        true,
-		MaxSize:        DefaultMaxSize,
-		MaxAge:         toml.Duration(DefaultMaxAge),
-		RetryRateLimit: DefaultRetryRateLimit,
-		RetryInterval:  toml.Duration(DefaultRetryInterval),
+		Enabled:          true,
+		MaxSize:          DefaultMaxSize,
+		MaxAge:           toml.Duration(DefaultMaxAge),
+		RetryRateLimit:   DefaultRetryRateLimit,
+		RetryInterval:    toml.Duration(DefaultRetryInterval),
+		RetryMaxInterval: toml.Duration(DefaultRetryMaxInterval),
 	}
 }
diff --git a/services/hh/config_test.go b/services/hh/config_test.go
index 2d2f4f217e..9963a31af3 100644
--- a/services/hh/config_test.go
+++ b/services/hh/config_test.go
@@ -14,6 +14,7 @@ func TestConfigParse(t *testing.T) {
 	if _, err := toml.Decode(`
 enabled = false
 retry-interval = "10m"
+retry-max-interval = "100m"
 max-size=2048
 max-age="20m"
 retry-rate-limit=1000
@@ -30,6 +31,10 @@ retry-rate-limit=1000
 		t.Fatalf("unexpected retry interval: got %v, exp %v", c.RetryInterval, exp)
 	}
 
+	if exp := 100 * time.Minute; c.RetryMaxInterval.String() != exp.String() {
+		t.Fatalf("unexpected retry max interval: got %v, exp %v", c.RetryMaxInterval, exp)
+	}
+
 	if exp := 20 * time.Minute; c.MaxAge.String() != exp.String() {
 		t.Fatalf("unexpected max age: got %v, exp %v", c.MaxAge, exp)
 	}
diff --git a/services/hh/service.go b/services/hh/service.go
index cf9424a123..88860350eb 100644
--- a/services/hh/service.go
+++ b/services/hh/service.go
@@ -119,17 +119,29 @@ func (s *Service) WriteShard(shardID, ownerID uint64, points []models.Point) err
 
 func (s *Service) retryWrites() {
 	defer s.wg.Done()
-	ticker := time.NewTicker(time.Duration(s.cfg.RetryInterval))
-	defer ticker.Stop()
+	currInterval := time.Duration(s.cfg.RetryInterval)
+	if currInterval > time.Duration(s.cfg.RetryMaxInterval) {
+		currInterval = time.Duration(s.cfg.RetryMaxInterval)
+	}
+
 	for {
+
 		select {
 		case <-s.closing:
 			return
-		case <-ticker.C:
+		case <-time.After(currInterval):
 			s.statMap.Add(processReq, 1)
 			if err := s.HintedHandoff.Process(); err != nil && err != io.EOF {
 				s.statMap.Add(processReqFail, 1)
 				s.Logger.Printf("retried write failed: %v", err)
+
+				currInterval = currInterval * 2
+				if currInterval > time.Duration(s.cfg.RetryMaxInterval) {
+					currInterval = time.Duration(s.cfg.RetryMaxInterval)
+				}
+			} else {
+				// Success! Return to configured interval.
+				currInterval = time.Duration(s.cfg.RetryInterval)
 			}
 		}
 	}