fix: randomise the initial grace period to avoid collisions

The previous algorithm was using binary exponential-backoff with
a +- 10% jitter to calculate the grace period.
Because there can be multiple lambda environments we need to
mitigate collisions:

We cannot use 0 as the first delay because functions failing closer to
each other will collide. The issue would then be propagated by the
small jitter for lower delays.

This change adds an initial delay of n seconds to the first reconnection
attempt.
n is randomly generated in a closed interval to account for collisions
while keeping in mind usability and user experience.

Author: kruskall

apm-lambda-extension/apmproxy/apmserver.go

@@ -197,6 +197,13 @@ func (c *Client) SetApmServerTransportState(ctx context.Context, status Status)
 
 // ComputeGracePeriod https://github.com/elastic/apm/blob/main/specs/agents/transport.md#transport-errors
 func (c *Client) ComputeGracePeriod() time.Duration {
+	// If reconnectionCount is 0, returns a random number in an interval.
+	// The grace period for the first reconnection count was 0 but that
+	// leads to collisions with multiple environments.
+	if c.ReconnectionCount == 0 {
+		gracePeriod := rand.Float64() * 5
+		return time.Duration(gracePeriod * float64(time.Second))
+	}
 	gracePeriodWithoutJitter := math.Pow(math.Min(float64(c.ReconnectionCount), 6), 2)
 	jitter := rand.Float64()/5 - 0.1
 	return time.Duration((gracePeriodWithoutJitter + jitter*gracePeriodWithoutJitter) * float64(time.Second))

apm-lambda-extension/apmproxy/apmserver_test.go

@@ -129,7 +129,7 @@ func TestGracePeriod(t *testing.T) {
 
 	apmClient.ReconnectionCount = 0
 	val0 := apmClient.ComputeGracePeriod().Seconds()
-	assert.Equal(t, val0, float64(0))
+	assert.LessOrEqual(t, val0, 5.0)
 
 	apmClient.ReconnectionCount = 1
 	val1 := apmClient.ComputeGracePeriod().Seconds()
@@ -192,7 +192,7 @@ func TestSetPendingTransport(t *testing.T) {
 	apmClient.SetApmServerTransportState(context.Background(), apmproxy.Failing)
 	require.Eventually(t, func() bool {
 		return apmClient.Status != apmproxy.Failing
-	}, 1*time.Second, 50*time.Millisecond)
+	}, 5*time.Second, 50*time.Millisecond)
 	assert.True(t, apmClient.Status == apmproxy.Pending)
 	assert.Equal(t, apmClient.ReconnectionCount, 0)
 }
@@ -313,7 +313,7 @@ func TestEnterBackoffFromFailing(t *testing.T) {
 	apmClient.SetApmServerTransportState(context.Background(), apmproxy.Failing)
 	require.Eventually(t, func() bool {
 		return apmClient.Status != apmproxy.Failing
-	}, 1*time.Second, 50*time.Millisecond)
+	}, 5*time.Second, 50*time.Millisecond)
 	assert.Equal(t, apmClient.Status, apmproxy.Pending)
 
 	assert.Error(t, apmClient.PostToApmServer(context.Background(), agentData))
@@ -366,7 +366,7 @@ func TestAPMServerRecovery(t *testing.T) {
 	apmClient.SetApmServerTransportState(context.Background(), apmproxy.Failing)
 	require.Eventually(t, func() bool {
 		return apmClient.Status != apmproxy.Failing
-	}, 1*time.Second, 50*time.Millisecond)
+	}, 5*time.Second, 50*time.Millisecond)
 	assert.Equal(t, apmClient.Status, apmproxy.Pending)
 
 	assert.NoError(t, apmClient.PostToApmServer(context.Background(), agentData))
@@ -410,7 +410,7 @@ func TestAPMServerAuthFails(t *testing.T) {
 	apmClient.SetApmServerTransportState(context.Background(), apmproxy.Failing)
 	require.Eventually(t, func() bool {
 		return apmClient.Status != apmproxy.Failing
-	}, 1*time.Second, 50*time.Millisecond)
+	}, 5*time.Second, 50*time.Millisecond)
 	assert.Equal(t, apmClient.Status, apmproxy.Pending)
 	assert.NoError(t, apmClient.PostToApmServer(context.Background(), agentData))
 	assert.NotEqual(t, apmClient.Status, apmproxy.Healthy)
@@ -458,7 +458,7 @@ func TestContinuedAPMServerFailure(t *testing.T) {
 	apmClient.SetApmServerTransportState(context.Background(), apmproxy.Failing)
 	require.Eventually(t, func() bool {
 		return apmClient.Status != apmproxy.Failing
-	}, 1*time.Second, 50*time.Millisecond)
+	}, 5*time.Second, 50*time.Millisecond)
 	assert.Equal(t, apmClient.Status, apmproxy.Pending)
 	assert.Error(t, apmClient.PostToApmServer(context.Background(), agentData))
 	assert.Equal(t, apmClient.Status, apmproxy.Failing)

apm-lambda-extension/apmproxy/client.go

@@ -20,6 +20,7 @@ package apmproxy
 import (
 	"bytes"
 	"errors"
+	"math/rand"
 	"net/http"
 	"strings"
 	"sync"
@@ -81,5 +82,7 @@ func NewClient(opts ...Option) (*Client, error) {
 		c.serverURL = c.serverURL + "/"
 	}
 
+	rand.Seed(time.Now().UnixNano())
+
 	return &c, nil
 }