Add exponential backoff to healthcheck (#3643)

Move exponential backoff initialization to Start()

Signed-off-by: RickyRajinder <singh.sangh@gmail.com>

Move comment

Increase max interval and update README

Remove trailing whitespace

Change Start() param name back to interval

go.mod

@@ -12,6 +12,7 @@ require (
 	github.com/apache/thrift v0.13.0 // indirect
 	github.com/aws/aws-sdk-go v1.28.9
 	github.com/caddyserver/caddy v1.0.4
+	github.com/cenkalti/backoff/v4 v4.0.0
 	github.com/coredns/federation v0.0.0-20190818181423-e032b096babe
 	github.com/coreos/go-systemd v0.0.0-20190212144455-93d5ec2c7f76 // indirect
 	github.com/coreos/pkg v0.0.0-20180928190104-399ea9e2e55f // indirect

go.sum

@@ -78,6 +78,8 @@ github.com/bgentry/speakeasy v0.1.0/go.mod h1:+zsyZBPWlz7T6j88CTgSN5bM796AkVf0kB
 github.com/caddyserver/caddy v1.0.4 h1:wwuGSkUHo6RZ3oMpeTt7J09WBB87X5o+IZN4dKehcQE=
 github.com/caddyserver/caddy v1.0.4/go.mod h1:uruyfVsyMcDb3IOzSKsi1x0wOjy1my/PxOSTcD+24jM=
 github.com/cenkalti/backoff/v3 v3.0.0/go.mod h1:cIeZDE3IrqwwJl6VUwCN6trj1oXrTS4rc0ij+ULvLYs=
+github.com/cenkalti/backoff/v4 v4.0.0 h1:6VeaLF9aI+MAUQ95106HwWzYZgJJpZ4stumjj6RFYAU=
+github.com/cenkalti/backoff/v4 v4.0.0/go.mod h1:eEew/i+1Q6OrCDZh3WiXYv3+nJwBASZ8Bog/87DQnVg=
 github.com/census-instrumentation/opencensus-proto v0.2.0/go.mod h1:f6KPmirojxKA12rnyqOA5BBL4O983OfeGPqjHWSTneU=
 github.com/census-instrumentation/opencensus-proto v0.2.1/go.mod h1:f6KPmirojxKA12rnyqOA5BBL4O983OfeGPqjHWSTneU=
 github.com/cespare/xxhash/v2 v2.1.1 h1:6MnRN8NT7+YBpUIWxHtefFZOKTAPgGjpQSxqLNn0+qY=

plugin/forward/README.md

@@ -9,8 +9,10 @@
 The *forward* plugin re-uses already opened sockets to the upstreams. It supports UDP, TCP and
 DNS-over-TLS and uses in band health checking.
 
-When it detects an error a health check is performed. This checks runs in a loop, every *0.5s*, for
-as long as the upstream reports unhealthy. Once healthy we stop health checking (until the next
+When it detects an error a health check is performed. This checks runs in a loop,
+starting with a *0.5s* interval and exponentially backing off with randomized intervals
+up to *60s* for as long as the upstream reports unhealthy. The exponential backoff
+will reset to *0.5s* after 15 minutes. Once healthy we stop health checking (until the next
 error). The health checks use a recursive DNS query (`. IN NS`) to get upstream health. Any response
 that is not a network error (REFUSED, NOTIMPL, SERVFAIL, etc) is taken as a healthy upstream. The
 health check uses the same protocol as specified in **TO**. If `max_fails` is set to 0, no checking

plugin/pkg/up/up.go

@@ -5,6 +5,8 @@ package up
 import (
 	"sync"
 	"time"
+
+	"github.com/cenkalti/backoff/v4"
 )
 
 // Probe is used to run a single Func until it returns true (indicating a target is healthy). If an Func
@@ -13,8 +15,7 @@ import (
 type Probe struct {
 	sync.Mutex
 	inprogress int
-	interval   time.Duration
-	max        time.Duration
+	expBackoff backoff.ExponentialBackOff
 }
 
 // Func is used to determine if a target is alive. If so this function must return nil.
@@ -31,7 +32,13 @@ func (p *Probe) Do(f Func) {
 		return
 	}
 	p.inprogress = active
-	interval := p.interval
+	interval := p.expBackoff.NextBackOff()
+	// If exponential backoff has reached the maximum elapsed time (15 minutes),
+	// reset it and try again
+	if interval == -1 {
+		p.expBackoff.Reset()
+		interval = p.expBackoff.NextBackOff()
+	}
 	p.Unlock()
 	// Passed the lock. Now run f for as long it returns false. If a true is returned
 	// we return from the goroutine and we can accept another Func to run.
@@ -42,9 +49,6 @@ func (p *Probe) Do(f Func) {
 				break
 			}
 			time.Sleep(interval)
-			if i%2 == 0 && i < 4 { // 4 is 2 doubles, so no need to increase anymore - this is *also* checked in double()
-				p.double()
-			}
 			p.Lock()
 			if p.inprogress == stop {
 				p.Unlock()
@@ -60,15 +64,6 @@ func (p *Probe) Do(f Func) {
 	}()
 }
 
-func (p *Probe) double() {
-	p.Lock()
-	p.interval *= 2
-	if p.interval > p.max {
-		p.interval = p.max
-	}
-	p.Unlock()
-}
-
 // Stop stops the probing.
 func (p *Probe) Stop() {
 	p.Lock()
@@ -77,10 +72,20 @@ func (p *Probe) Stop() {
 }
 
 // Start will initialize the probe manager, after which probes can be initiated with Do.
+// Initializes exponential backoff using the given interval duration
 func (p *Probe) Start(interval time.Duration) {
 	p.Lock()
-	p.interval = interval
-	p.max = interval * multiplier
+	eB := &backoff.ExponentialBackOff{
+		InitialInterval:     interval,
+		RandomizationFactor: backoff.DefaultRandomizationFactor,
+		Multiplier:          backoff.DefaultMultiplier,
+		MaxInterval:         backoff.DefaultMaxInterval,
+		MaxElapsedTime:      backoff.DefaultMaxElapsedTime,
+		Stop:                backoff.Stop,
+		Clock:               backoff.SystemClock,
+	}
+	p.expBackoff = *eB
+	p.expBackoff.Reset()
 	p.Unlock()
 }
 
@@ -88,6 +93,4 @@ const (
 	idle = iota
 	active
 	stop
-
-	multiplier = 4
 )