fix(test): embed cluster-agent Helm values as testdata yaml, trim comments

Author: louis-cqrl

test/new-e2e/tests/agent-telemetry/errortracking_cluster_agent_test.go

@@ -7,6 +7,7 @@ package agenttelemetry
 
 import (
 	"context"
+	_ "embed"
 	"fmt"
 	"io"
 	"strings"
@@ -32,47 +33,28 @@ import (
 // the datadog-agent Helm release into.
 const clusterAgentDatadogNamespace = "datadog"
 
-// leaderElectionErrorMessage is logged by KubeASCheck.Run at ERROR level on
-// every check run (~15s, no rate limiter) once leader election is disabled —
-// a deterministic trigger for a binary that runs no Python checks.
+// leaderElectionErrorMessage is logged at ERROR on every check run once
+// leader election is disabled, a deterministic trigger for this suite.
 const leaderElectionErrorMessage = "Leader Election not enabled"
 
-// clusterAgentErrorTrackingEnabledHelmValues disables leader election (to
-// generate a repeating ERROR log) and enables the errortracking pipeline
-// with a fast flush so the wire-shape assertions below run quickly.
-const clusterAgentErrorTrackingEnabledHelmValues = `
-datadog:
-  leaderElection: false
-clusterAgent:
-  envDict:
-    DD_AGENT_TELEMETRY_ENABLED: "true"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_ENABLED: "true"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_FLUSH_INTERVAL_SECONDS: "1"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_BOUNCER_WINDOW_SECONDS: "0"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_STARTUP_JITTER_SECONDS: "0"
-`
+// clusterAgentErrorTrackingEnabledHelmValues disables leader election and
+// enables the errortracking pipeline with a fast flush.
+//
+//go:embed testdata/errortracking-cluster-agent-enabled.yaml
+var clusterAgentErrorTrackingEnabledHelmValues string
 
 // clusterAgentErrorTrackingDisabledHelmValues mirrors the enabled config but
-// omits errortracking.enabled, which defaults to false, while still forcing
-// the leader-election error so the negative assertion is meaningful.
-const clusterAgentErrorTrackingDisabledHelmValues = `
-datadog:
-  leaderElection: false
-clusterAgent:
-  envDict:
-    DD_AGENT_TELEMETRY_ENABLED: "true"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_FLUSH_INTERVAL_SECONDS: "1"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_BOUNCER_WINDOW_SECONDS: "0"
-    DD_AGENT_TELEMETRY_ERRORTRACKING_STARTUP_JITTER_SECONDS: "0"
-`
+// omits errortracking.enabled, which defaults to false.
+//
+//go:embed testdata/errortracking-cluster-agent-disabled.yaml
+var clusterAgentErrorTrackingDisabledHelmValues string
 
 type errorTrackingClusterAgentSuite struct {
 	e2e.BaseSuite[environments.Kubernetes]
 }
 
 // TestErrorTrackingClusterAgentSuite is the cluster-agent variant of
-// TestAgentTelemetryErrorTrackingSuite, exercising the same
-// pkg/util/log/errortracking → comp/core/agenttelemetry pipeline.
+// TestAgentTelemetryErrorTrackingSuite.
 func TestErrorTrackingClusterAgentSuite(t *testing.T) {
 	e2e.Run(t, &errorTrackingClusterAgentSuite{},
 		e2e.WithProvisioner(provkind.Provisioner(
@@ -86,11 +68,7 @@ func TestErrorTrackingClusterAgentSuite(t *testing.T) {
 }
 
 // getClusterAgentPodNames returns the names of all running cluster-agent
-// pods. This suite's Helm deployment runs clusterAgent.replicas: 2 for HA
-// (see test/e2e-framework's kindvm base values), so callers must not assume
-// a single pod. The leader-election error asserted on below is logged by
-// each replica's own local corecheck runner independently of DCA-level
-// leader election, so any one replica emitting it is sufficient.
+// pods. This suite runs clusterAgent.replicas: 2, so callers must not assume a single pod.
 func (s *errorTrackingClusterAgentSuite) getClusterAgentPodNames() []string {
 	t := s.T()
 	pods, err := s.Env().KubernetesCluster.Client().CoreV1().Pods(clusterAgentDatadogNamespace).List(t.Context(), metav1.ListOptions{
@@ -106,12 +84,7 @@ func (s *errorTrackingClusterAgentSuite) getClusterAgentPodNames() []string {
 }
 
 // getClusterAgentContainerLogs returns the "cluster-agent" container's stdout
-// for podName, restricted to entries logged at or after since. The Kubelet's
-// log endpoint filters this server-side, so unlike a PodExec'd file read there
-// is no local file, offset, or rolling-writer state to reconcile: cluster-agent
-// runs containerized (via this suite's kind provisioner) and, like every other
-// binary in a Helm-deployed pod, relies on the container runtime to capture its
-// stdout rather than writing a log file to disk.
+// for podName, restricted to entries logged at or after since.
 func (s *errorTrackingClusterAgentSuite) getClusterAgentContainerLogs(ctx context.Context, podName string, since time.Time) (string, error) {
 	sinceTime := metav1.NewTime(since)
 	stream, err := s.Env().KubernetesCluster.Client().CoreV1().Pods(clusterAgentDatadogNamespace).GetLogs(podName, &corev1.PodLogOptions{
@@ -127,12 +100,8 @@ func (s *errorTrackingClusterAgentSuite) getClusterAgentContainerLogs(ctx contex
 }
 
 // TestPayloadShape verifies the cluster-agent's own leader-election-gated
-// ERROR log reaches FakeIntake with the expected wire shape and an
-// agent.flavor tag identifying the emitter as cluster_agent rather than agent.
+// ERROR log reaches FakeIntake with an agent.flavor tag of cluster_agent.
 func (s *errorTrackingClusterAgentSuite) TestPayloadShape() {
-	// BeforeTest already reset the environment to the suite's original
-	// (enabled) provisioner regardless of run order, and the leader-election
-	// error recurs on every check run, so no re-provisioning is needed here.
 	require.NoError(s.T(), s.Env().FakeIntake.Client().FlushServerAndResetAggregators())
 
 	var logs []*aggregator.AgentTelemetryLog
@@ -148,9 +117,8 @@ func (s *errorTrackingClusterAgentSuite) TestPayloadShape() {
 	}
 }
 
-// TestDisabledByDefault verifies that when the errortracking stanza omits
-// `enabled` (defaulting to false), no agent-logs records reach FakeIntake even
-// though the leader-election error keeps firing locally.
+// TestDisabledByDefault verifies that when errortracking omits `enabled`
+// (defaulting to false), no agent-logs records reach FakeIntake.
 func (s *errorTrackingClusterAgentSuite) TestDisabledByDefault() {
 	s.UpdateEnv(provkind.Provisioner(
 		provkind.WithRunOptions(
@@ -165,17 +133,8 @@ func (s *errorTrackingClusterAgentSuite) TestDisabledByDefault() {
 	podNames := s.getClusterAgentPodNames()
 	since := time.Now()
 
-	// Wait until the leader-election error appears in at least one replica's own
-	// stdout, confirming the error is generated locally before asserting it is
-	// not forwarded to telemetry. Only the elected DCA replica is guaranteed to
-	// run the check that emits it, so all replicas are polled and any one of
-	// them containing the message satisfies the wait.
-	//
-	// The window is generous (matching TestPayloadShape's FakeIntake wait)
-	// because this runs right after a Helm upgrade rolls the deployment: on a
-	// resource-constrained CI node, the new pods' corecheck scheduler can take
-	// longer to get its first tick in than the default-check interval alone
-	// would suggest.
+	// Wait until the leader-election error appears in at least one replica's
+	// stdout before asserting it is not forwarded to telemetry.
 	ok := assert.EventuallyWithT(s.T(), func(c *assert.CollectT) {
 		statuses := make([]string, 0, len(podNames))
 		for _, podName := range podNames {
@@ -192,8 +151,7 @@ func (s *errorTrackingClusterAgentSuite) TestDisabledByDefault() {
 		assert.Fail(c, "leader-election error not yet found in any cluster-agent replica's log: "+strings.Join(statuses, "; "))
 	}, 2*time.Minute, 5*time.Second, "timed out waiting for leader-election error to appear in cluster-agent log")
 
-	// On failure, dump each replica's actual log tail so a re-run shows what the
-	// check runner was doing instead of just a stale/zero count.
+	// On failure, dump each replica's log tail for debugging.
 	if !ok {
 		for _, podName := range podNames {
 			out, err := s.getClusterAgentContainerLogs(ctx, podName, since)
@@ -206,9 +164,7 @@ func (s *errorTrackingClusterAgentSuite) TestDisabledByDefault() {
 		s.T().FailNow()
 	}
 
-	// Confirm nothing is forwarded. The config sets flush_interval_seconds: 1, so
-	// 5 s covers five flush cycles: if a regression enabled the forwarder, it would
-	// flush within this window and the assertion would catch it.
+	// Confirm nothing is forwarded across five flush cycles (flush_interval_seconds: 1).
 	assert.Never(s.T(), func() bool {
 		logs, err := s.Env().FakeIntake.Client().GetAgentTelemetryLogs()
 		require.NoError(s.T(), err)

test/new-e2e/tests/agent-telemetry/testdata/errortracking-cluster-agent-disabled.yaml

@@ -0,0 +1,8 @@
+datadog:
+  leaderElection: false
+clusterAgent:
+  envDict:
+    DD_AGENT_TELEMETRY_ENABLED: "true"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_FLUSH_INTERVAL_SECONDS: "1"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_BOUNCER_WINDOW_SECONDS: "0"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_STARTUP_JITTER_SECONDS: "0"

test/new-e2e/tests/agent-telemetry/testdata/errortracking-cluster-agent-enabled.yaml

@@ -0,0 +1,9 @@
+datadog:
+  leaderElection: false
+clusterAgent:
+  envDict:
+    DD_AGENT_TELEMETRY_ENABLED: "true"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_ENABLED: "true"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_FLUSH_INTERVAL_SECONDS: "1"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_BOUNCER_WINDOW_SECONDS: "0"
+    DD_AGENT_TELEMETRY_ERRORTRACKING_STARTUP_JITTER_SECONDS: "0"