Fix flaky integration tests: increase e2e timeouts and retry limits

integration/e2e/composite_service.go

@@ -25,7 +25,7 @@ func NewCompositeHTTPService(services ...*HTTPService) *CompositeHTTPService {
 		retryBackoff: backoff.New(context.Background(), backoff.Config{
 			MinBackoff: 300 * time.Millisecond,
 			MaxBackoff: 600 * time.Millisecond,
-			MaxRetries: 50, // Sometimes the CI is slow ¯\_(ツ)_/¯
+			MaxRetries: 100, // Sometimes the CI is slow ¯\_(ツ)_/¯
 		}),
 	}
 }

integration/e2e/service.go

@@ -71,7 +71,7 @@ func NewConcreteService(
 		retryBackoff: backoff.New(context.Background(), backoff.Config{
 			MinBackoff: 300 * time.Millisecond,
 			MaxBackoff: 600 * time.Millisecond,
-			MaxRetries: 50, // Sometimes the CI is slow ¯\_(ツ)_/¯
+			MaxRetries: 100, // Sometimes the CI is slow ¯\_(ツ)_/¯
 		}),
 	}
 }
@@ -476,7 +476,7 @@ func (p *TCPReadinessProbe) Ready(service *ConcreteService) (err error) {
 		return errors.New("service has stopped")
 	}
 
-	conn, err := net.DialTimeout("tcp", endpoint, time.Second)
+	conn, err := net.DialTimeout("tcp", endpoint, 5*time.Second)
 	if err != nil {
 		return err
 	}

integration/e2e/util.go

@@ -84,14 +84,14 @@ func BuildArgs(flags map[string]string) []string {
 }
 
 func GetRequest(url string) (*http.Response, error) {
-	const timeout = 1 * time.Second
+	const timeout = 5 * time.Second
 
 	client := &http.Client{Timeout: timeout}
 	return client.Get(url)
 }
 
 func PostRequest(url string) (*http.Response, error) {
-	const timeout = 1 * time.Second
+	const timeout = 5 * time.Second
 
 	client := &http.Client{Timeout: timeout}
 	return client.Post(url, "", strings.NewReader(""))