From fc89ea42ee0d60cd8f6c31496c743c1c441e8792 Mon Sep 17 00:00:00 2001
From: zirain <zirain2009@gmail.com>
Date: Thu, 11 Jul 2024 19:56:09 +0800
Subject: [PATCH] e2e: make sure ALS server is ready (#3816)

Signed-off-by: zirain <zirain2009@gmail.com>
---
 test/e2e/tests/accesslog.go | 23 +++++++++++++++++++++--
 test/e2e/tests/utils.go     |  8 ++++----
 2 files changed, 25 insertions(+), 6 deletions(-)

diff --git a/test/e2e/tests/accesslog.go b/test/e2e/tests/accesslog.go
index b0de4543cb7f..1fecc669c340 100644
--- a/test/e2e/tests/accesslog.go
+++ b/test/e2e/tests/accesslog.go
@@ -152,7 +152,22 @@ var ALSTest = suite.ConformanceTest{
 			routeNN := types.NamespacedName{Name: "accesslog-als", Namespace: ns}
 			gwNN := types.NamespacedName{Name: "same-namespace", Namespace: ns}
 			gwAddr := kubernetes.GatewayAndHTTPRoutesMustBeAccepted(t, suite.Client, suite.TimeoutConfig, suite.ControllerName, kubernetes.NewGatewayRef(gwNN), routeNN)
-			preCount := ALSLogCount(t, suite)
+
+			preCount := 0
+			// make sure ALS server metric endpoint is ready
+			if err := wait.PollUntilContextTimeout(context.TODO(), time.Second, time.Minute, true,
+				func(ctx context.Context) (bool, error) {
+					curCount, err := ALSLogCount(suite)
+					if err != nil {
+						t.Logf("failed to get log count from loki: %v", err)
+						return false, nil
+					}
+					preCount = curCount
+					return true, nil
+				}); err != nil {
+				t.Errorf("failed to get log count from loki: %v", err)
+			}
+
 			expectedResponse := httputils.ExpectedResponse{
 				Request: httputils.Request{
 					Path: "/als",
@@ -170,7 +185,11 @@ var ALSTest = suite.ConformanceTest{
 
 			if err := wait.PollUntilContextTimeout(context.TODO(), time.Second, time.Minute, true,
 				func(ctx context.Context) (bool, error) {
-					curCount := ALSLogCount(t, suite)
+					curCount, err := ALSLogCount(suite)
+					if err != nil {
+						t.Logf("failed to get log count from loki: %v", err)
+						return false, nil
+					}
 					return preCount < curCount, nil
 				}); err != nil {
 				t.Errorf("failed to get log count from loki: %v", err)
diff --git a/test/e2e/tests/utils.go b/test/e2e/tests/utils.go
index b29d2b48d0d3..4ee2dbebcf7f 100644
--- a/test/e2e/tests/utils.go
+++ b/test/e2e/tests/utils.go
@@ -369,18 +369,18 @@ func WaitForLoadBalancerAddress(t *testing.T, client client.Client, timeout time
 	return ipAddr, nil
 }
 
-func ALSLogCount(t *testing.T, suite *suite.ConformanceTestSuite) int {
+func ALSLogCount(suite *suite.ConformanceTestSuite) (int, error) {
 	metricPath, err := RetrieveURL(suite.Client, types.NamespacedName{
 		Namespace: "monitoring",
 		Name:      "envoy-als",
 	}, 19001, "/metrics")
 	if err != nil {
-		t.Fatalf("failed to get metric url: %v", err)
+		return -1, err
 	}
 
 	countMetric, err := RetrieveMetric(metricPath, "log_count", time.Second)
 	if err != nil {
-		t.Fatalf("failed to get metric: %v", err)
+		return -1, err
 	}
 
 	total := 0
@@ -390,7 +390,7 @@ func ALSLogCount(t *testing.T, suite *suite.ConformanceTestSuite) int {
 		}
 	}
 
-	return total
+	return total, nil
 }
 
 // QueryLogCountFromLoki queries log count from loki